The Remote Sensing of Ancestral Puebloan Sites by David Eugene Witt February 1, 2010 A project submitted to the Faculty of the Graduate School of the State University of New York at Buffalo in partial fulfillment of the requirements for the degree of Master of Arts Department of Geography Copyright by David Eugene Witt 2010 ii Dedication To Mom & Dad. Acknowledgements I would like to thank my advisor, Dr. Sean Bennett, and my second reader, Dr. Le Wang, for their unfailing patience and flexibility. I would also like to thank Linda Wheelburger, principal investigator of the Totah Archaeological Project, for her aid in obtaining archaeological site locations within the Middle San Juan River region. iii Table of Contents Dedication & Acknowledgements iii Table of Figures vi Abstract vii Introduction 1 Objectives 3 Geographical & Environmental Overview of the Colorado Plateau 5 The Modern Geography & Environment 5 The Historical Environment 7 Historical Overview of the Middle San Juan Region 11 Archaic (pre-1200 BC) 12 Basketmaker II (1200 BC – AD 500) 12 Basketmaker III (AD 500 – AD 750) 14 Pueblo I (AD 750 – AD 900) 14 Pueblo II (AD 900 – AD 1150) 15 Pueblo III (AD 1150 – AD 1350) 16 Pueblo IV (AD 1350 – AD 1600) 17 Historical Review of Remote Sensing Techniques 18 Aerial Photography 18 Electric Resistivity & Electromagnetic Surveying 21 Magnetometry 23 Ground Penetrating Radar & Sonar 24 iv Analytical Chemistry Techniques 25 Satellite Photography 27 Application of Remote Sensing Techniques to the Archaeology of the Southwest Literature Review 31 33 ―The Use of ASTER Satellite Imagery in Archaeological Contexts‖ 33 ―Classification of Multispectral ASTER Imagery in Archaeological Settlement Survey in the Near East‖ 35 ―Archaeological Site Detection: The Importance of Contrast‖ 36 Discussion 37 Methodology 39 Results & Conclusion 48 Works Cited 49 v Table of Figures Figure 1: Map of Project Area 2 Figure 2: Map of Colorado Plateau 6 Figure 3: ASTER Image of Project Area, Bands 1, 2, 3 41 Figure 4: Eigenvalue Graphs, Principal Component Analyses on VNIR and SWIR 42 Figure 5: PCA Image, VNIR Bands 2, 3, SWIR Band 1 43 Figure 6: Unsupervised Isodata Reclassification 44 Figure 7: Location of Archaeological Sites 45 Figure 8: Maximum Likelihood Reclassification, on Principal Component Analysis of NVIR Bands vi 47 Abstract This paper is the result of a master‗s project to incorporate the use of imagery obtained by the ASTER space-borne instrument to aid in the detection of surface archaeological sites along the San Juan River of northwestern New Mexico and the surrounding area within the Four Corners region of the American Southwest. The methodologies used various unsupervised (IsoData) and supervised (maximum likelihood, minimum distance, and spectral information divergence) classification methods. These methods were attempted, and subsequently compared, to determine if any method, or a combination of the methods, would provide more accurate results within the project area. If these methods determined any potential archaeological site locations within the study area, these locations would have been communicated to Linda Wheelburger, principal investigator of the Totah Archaeological Project, for field identification. Unfortunately, these methods did not result any in usable data, and no potential site locations were identified. vii I. Introduction Archaeology is concerned with the detection, recovery, and conservation of artifacts and features, material remains of past cultures which aid us in understanding the nature of those past cultures. Artifacts are portable, such as fragments of pottery or glass, or stone tools. On the other hand, features are non-portable, such as walls, building foundations, and alterations of the ground and soil. Traditionally, the detection of these materials has been accomplished through methods such as field walking, surveying, and random luck, but remote sensing techniques have been increasingly used by archaeologists as a method to focus their efforts by identifying areas with a higher likelihood of containing artifacts or by identifying features in the soil before surveying or excavation has even begun. These techniques involve the incorporation of a variety of near-earth, aerial-borne, and space-borne instruments. This paper is the result of a master‘s project to incorporate the use of imagery obtained by the ASTER space-borne instrument to aid in the detection of surface archaeological sites along the San Juan River of northwestern New Mexico and the surrounding area within the Four Corners region of the American Southwest. As such, it will first address the objectives of the project. Afterwards, it will provide a brief geographical, environmental, and historical overview of the study area, and then a review of the application of remote sensing techniques for archaeological purposes. A more detailed literature review of selected case examples will follow, proceeded by a discussion of the project methodology and results. Finally, a concluding discussion and remarks will be addressed. The study area of the project is located in the Middle San Juan Region and is centered on the city of Farmington, New Mexico, located at 36° 45‘ 6‖ North Latitude and 108° 11‘ 23‖ West Longitude. This is located within the Colorado Plateau, in Universal Transverse Mercator 1 (UTM) Zone 12N, with a northing value of approximately 4,071,000 meters and an easting value of approximately 750,900 meters from the southwest corner of the zone. The datum used during the study was North American Datum 1927 (NAD83). This is because the archaeological sites were recorded using the North America Datum 1927 (NAD27) as is required by New Mexican state law. These differing datums were taken into account during the project undertaking. The majority of the sites already known in the project area date to the Pueblo II to Pueblo III periods (AD 900 – AD 1350), though sites dating to previous and later periods have been discovered. Due to the dates of the known sites in the project area, any sites discovered by the project will be assumed to date from the same periods and have the same cultural affinity. This assumption, however, will need to be tested by future field work. 2 II. Objectives The primary objective of the project was to develop and test a regionalized methodology for employing remotely sensed data in the form of ASTER imagery for the classification and detection of archaeological sites throughout the project area. This was accomplished through the use of two computer software suites, ITT Visual Information Solutions‘ ENVI 4.4 and Environmental Systems Research Institute‘s (ESRI) ArcGIS 9.3. ENVI 4.4 was used to conduct various analyses and post-processing techniques upon multispectral images, whereas ArcGIS 9.3 was used to manipulate and synthesize geographic data to highlight areas of interest which might contain archaeological sites. The methodologies incorporated include various unsupervised (IsoData) and supervised (maximum likelihood, minimum distance, and spectral information divergence) classification methods. These methods were attempted, and subsequently compared, to determine if any method, or a combination of the methods, would provide more accurate results within the project area. The methods were initially developed by several computer scientists and archaeologists, and the development of these methods will be discussed in depth in Chapter VII. Unfortunately, the methodologies adapted from these papers were not successful for this project. Secondary objectives of this project were to produce a list of potential archaeological site locations within the project area for comparison and incorporation with the records of the San Juan Museum based at Salmon Ruins, New Mexico and the Totah Archaeological Project of San Juan College. These sites were then to be investigated in the field to determine if the methodology had correctly identified the sites. However, due to the unsuccessful nature of the project, no sites were determined from analyzing the ASTER imagery and therefore it is impossible to create this list. Also, the results of this project were to be delivered, as a conference 3 poster, at the 75th Annual Meeting of the Society for American Archaeology, to be held April 14th to April 18th, 2010. These results are still planned to be delivered, and it is hoped that future work will refine a methodology that may help determine site location. Finally, because no new sites were determined, it was deemed unnecessary to provide the precise location of previously known archaeological sites. The presence and locations of these sites are considered privileged and confidential information. 4 III. Geographical & Environmental Overview of the Colorado Plateau This chapter provides a brief overview of the geography and environment (modern and historical) of the Colorado Plateau and the Middle San Juan Region. This is to familiarize the reader with the geographical conditions in which the project was set, and because the geography and environmental conditions influenced the development of the various cultures and the settlement patterns investigated by the project. Sources of water and arable land were, and still are, the most important resources available in the project area, and it is believed that any sites discovered by this project will be located near the former locations of those two resources. The historical environment as reconstructed will be discussed with evidence obtained from archaeological research conducted throughout the region. A. The Modern Geography & Environment The Colorado Plateau is a physiographic region within the intermontane plateaus of the southern Rockies, covering an area of approximately 337,000 square kilometers (130,000 square miles) in western Colorado, Utah, northern Arizona, and northwestern New Mexico. It is bounded on the northeast by Rocky Mountains in Colorado, and on the northwest by the Uinta and Wasatch Mountains in Utah. The southeast edge is demarcated by the Rio Grande Rift, and the southern by the Mogollon Rim. About 90% of the Plateau is drained by the Colorado River and its tributaries, such as the Muddy, Virgin, San Juan, and Little Colorado Rivers (Larson & Michaelson 1990: 229). 5 The modern environment of the Colorado Plateau is primarily one of strong zonation, which is the division of the landscape into various microclimates based on elevation, temperature, and precipitation. These different vegetation zones are the result of the presence of microclimates, with the primary factor being elevation, but the degree of slope and exposure to prevalent winds and sunshine are also factors. There are six zones, determined primarily by vegetation: sage desert, short grass, pinyon-juniper woodlands, pine parkland, mixed coniferous forests, and spruce-fir forests (Hevly 1988: 93). Elevation can range from 150 meters to 3500 meters, and the effect of elevation is such that with each 333 meter (1000 foot) increase in elevation, temperature is decreased by 2.5 degrees Celsius (Larson & Michaelson 1990: 229; Hevly 1988: 93). Along with the decrease in temperature, precipitation increases, with the distribution of precipitation influenced by 6 exposure, winds, and the presence of mountains, which creates rain shadows (Hevly 1988: 93). Due to the combination of these factors, it is hard to determine exactly which factors influence growth patterns, and to what extent they influence the patterns. The area is generally hot and arid, however considerable variability based on spatial, annual, and seasonal diversity exists within the region (Dean 1988: 122). The region is known to experience extreme cold, and temperatures below -18 degrees Celsius (0 degrees Fahrenheit) are occasionally the result of cold air masses moving south from Canada during the winter (Shepherd et al. 1999). Rainfall occurs during two seasons, July to September and December to March, and most of the average annual rainfall of 322 millimeters (12.7 inches) occurs during these two periods (Shepherd et al. 1999). The remaining months comprise two dry seasons, with the dryer climate occurring during the April to June season (Hevly 1988: 95). The seasonal nature of rainfall in the region is primarily due to the region‘s location between two sources of atmospheric moisture, the Gulf of Mexico to the east and the Pacific Ocean to the west. The thunderstorms that occur during the summer months are a result of air currents from the two bodies interacting with dryer air over the continent (Dean 1988: 122). The winter rains, however, originate from large frontal storms that originate from the northern Pacific Ocean and move eastward over the continent. A third, rarer source of rainfall is hurricanes which affect the western coast of Mexico, forcing warm moist air over the Colorado Plateau. The result of these rain patterns is the unique bimodal precipitation pattern with two rainy seasons and two dry seasons (Dean 1988: 123). B. The Historical Environment The regional geography has remained relatively stable over the past several millennia; however, one major exception was a series of volcanic eruptions located north of Flagstaff, 7 Arizona. These eruptions, occurring between AD 1080 and AD 1150, built the 340 meter tall volcanic cone now known as Sunset Crater. While the eruptions did have an effect on the local Sinagua population, it is not known to have effected other populations in the Southwest (Berlin et al. 1990). The past environmental differed greatly, though. While the climate experienced the same bimodal precipitation pattern previously discussed, the environmental history of the region indicates shits from wetter to dryer climates over the past several thousand years. The late Archaic/pre-Ancestral Puebloan period (approximately 1800 BC to 250 BC) was characterized by a climate slightly dryer than the modern climate (Hall 1988: 587). As such, this corresponds to an increase in relative levels of grassy, non-arboreal pollen levels throughout lower elevations in the San Juan Basin and elsewhere in the Colorado Plateau. An absence of pinyon and juniper in charcoal samples from the lower elevations during this time period likely illustrates an absence of those species from the area, as they were both considered excellent fuel sources and unlikely to be overlooked (Hall 1988: 588). Beginning around 250 BC, pinyon-juniper pollen levels increased to one-third that of modern levels, indicating a cooler, wetter environment (Hall 1988: 588). Pollen obtained from rodent middens throughout Chaco Canyon illustrate that pinyon pine numbers increased from AD 1 to AD 200 in that area to twice the modern levels, which would have been the result of a change in climate to a wetter environment in that particular area (Hall 1988: 588). Nevertheless, the increase of arboreal pollen in samples taken from throughout the Colorado Plateau indicates a shift in the climate throughout the region. Pollen records dating from AD 600 to AD 1000 indicate a shift back to dryer conditions throughout the Colorado Plateau, with samples originating from Chaco Canyon, southwest 8 Colorado, and the Virgin River in Utah (Hall 1988: 588, Larson & Michaelson 1990: 237, Kohler & Matthews 1988: 552). This dry spell continued until AD 800 when it intensified, peaked around AD 900, and finally ended circa AD 1000 (Kohler & Matthews 1988: 552). Throughout the plateau, pollen levels indicated a shift back to non-arboreal species, and the environment shifted to desert shrub grassland similar to the modern environment (Hall 1988: 588). This shift to dryer conditions would have influenced the location and growth of communities, which would have been located near larger rivers and other perennial sources of water. After AD 1000, moister conditions developed, as evidenced by growth rings, and this settlement pattern reversed (Hall 1988: 590). The conditions resulted with an increase in groundwater supplies, declining erosion (especially erosion of agricultural land), and increased precipitation (Gumerman & Gell-Mann 1994: 21). These, in turn, influenced ―the development of new traditions and the fluorescence of already established ones,‖ population increases throughout the Four Corners region, the development of large regional systems such as the Chacoan system and the establishment of permanent settlements in areas previously used only seasonally (Gumerman & Gell-Mann 1994: 21). Again, a drought developed in AD 1130 and lasted for 50 years, as illustrated by dendroclimatology. This drought is thought to have been a major contributing factor to the decline and collapse of the Chacoan system circa AD 1150 (Fagan 2000: 328-329). A return to moister conditions allowed for the resurgence of settlement construction throughout the Middle San Juan Region, but again this was not a permanent situation. A ―Great Drought‖ began circa AD 1275 and continued for 25 years. This placed enormous amounts of environmental stress on the existing populations. Once again, population centers throughout the 9 region experienced decline; in an extreme example the Entire San Juan Basin, including Mesa Verde on its northern periphery, was abandoned (Fagan 2000: 338). Circa AD 1350, which corresponds to the end of the Pueblo III period and the traditional date of the Ancestral Puebloan collapse, pinyon-juniper pollen levels increased in Chaco Canyon and along the San Juan River, as well at higher elevations such as on Chacra Mesa. These probably indicate reestablishment of pinyon-juniper forests following the Ancestral Puebloan collapse (Hall 1988: 589). These forests were descendent from small isolated stands, perhaps located along cliff walls (Hall 1988: 591). These stands served as sources of timber and fuel for the Ancestral Puebloans, and the reestablishment of these species indicate a complete and prolong abandonment of the areas. (Hall 1988: 591). 10 IV. Historical Overview of the Middle San Juan Region The cultural history of the Southwest, and the Middle San Juan Region specifically, is a long and complicated affair; while only a few cultures evolved in the area (this project focuses on one), numerous phases of cultural growth and realignments occurred. As such, a full recounting of the cultural history of the region is outside the scale of this project, and this chapter serves only as a brief introduction to the history of the region to allow the reader to understand terms used throughout the project report. It is arranged chronologically, and divided according to the traditional Pecos Classification scheme generally used throughout the Ancestral Puebloan area. Due to the imprecise nature of the Pecos Classification, the dates should be regarded as approximations. Currently known archaeological sites already located within the project area are limited to the Basketmaker III to Pueblo III periods (AD 500 – AD 1350), but it is possible that sites from the Archaic to the Pueblo IV periods may be discovered by this project. Because of this, the chapter will discuss the history of the region from the Archaic to the Pueblo IV (circa 1200 BC to AD 1600). Table 1: Pecos Classification Archaic Period Pre 1200 BC Basketmaker II Period 1200 BC – AD 500 Basketmaker III Period AD 500 – AD 750 Pueblo I Period AD 750 – AD 900 Pueblo II Period AD 900 – AD 1150 Pueblo III Period AD 1150 – AD 1350 Pueblo IV Period AD 1350 – AD 1600 11 A. Archaic (pre-1200 BC) The American Southwest has been inhabited since the Archaic period (also known as the Paleoindian period), and many well-known sites, such as Blackwater Draw at Clovis, New Mexico (dating to 11,000 BC) and Folsom, New Mexico (dating to 9,000 BC) are located in the region. However, despite the long history of human occupation in the area, the predominate lifestyle of the inhabitants remained one of mobility and dispersion, with sustenance obtained by hunting and gathering activities until circa 1500 BC (Gumerman & Gell-Mann 1994: 17). It was approximately this time that maize horticulture was introduced in the region (most likely imported from Mesoamerican contacts to the south, but possibly independently developed), and while it would take several thousand years for agriculture to become the dominant source of food in the region, the introduction of horticultural activities altered the cultural trajectory of the people of the region (Gumerman & Gell-Mann 1994: 17). The development of horticulture precipitated a slow move away from a mobile lifestyle to one of sedentism based upon the control of geographical areas and their agricultural resources, with this sedentism resulted in the growth of population centers in the region, primarily in the Rio Grande Valley and in the southern Arizona desert (Gumerman & Gell-Mann 1994: 17). However, despite a relationship between small population groups and local resources, archaeological evidence indicates that artifact styles were similar throughout the greater Southwest region throughout this time. B. Basketmaker II (1200 BC – AD 500) It was not until circa AD 200 that the separate population concentrations began to diverge into the three classic Southwestern traditions, the Hohokom (in southern and central Arizona), the Mogollon (in the highlands of eastern Arizona and western New Mexico), and the Ancestral 12 Puebloan culture colloquially known as the Anasazi, located in the San Juan Basin and in the immediate Four Corners area and west along the Arizona-Utah border into southern Nevada (Gumerman & Gell-Mann 1994: 17). These traditions were derived from the diverging artifact and architectural styles, and to a degree settlement pattern and location, rather than a specific set of cultural behavior related to a specific cultural tradition (Gumerman & Gell-Mann 1994: 17). However, the continued growth of the importance of agriculture as a food source independent of hunting or gathering activities resulted in a loss of mobility and a degree of sedentism that was unknown among adjacent areas such as the Great Basin and the Great Plains (Gumerman & GellMann 1994: 17). This dependence on agriculture even grew to include beans and squash, which, when included with maize as part of a normal diet, contained all the amino acids and protein necessary to maintain a healthy diet (Fagan 2000: 308). Throughout the Ancestral Puebloan area, population continued to grow into the Sixth Century AD. Seasonal communities consisting of pithouses (semi-subterranean dwellings) and underground storage rooms were founded throughout the southern Colorado plateaus. The development of these communities indicated a greater investment in sedentism and agriculture, and the indigenous invention of several technologies, including the bow & arrow, the twohanded mano and the trough metate, ceramic pottery, and the domestication of the turkey are evidence of the continued growth in the importance of local agricultural activities rather than hunting and gathering techniques (Gumerman & Gell-Mann 1994: 18). This development in community structure was termed the ―Basketmaker II‖ culture (―Basketmaker‖ because the culture was one of the first to develop basketry, and ―II‖ because archaeologists at the time had erroneously predicted a ―simpler‖ sedentary culture that made baskets but had not yet developed pottery). 13 C. Basketmaker III (AD 500 – AD 750) From about AD 500 to AD 750, the Basketmaker II culture continued to evolve, and the artifacts and features associated with the cultural tradition took on stylistic attributes that would later be identified as early Ancestral Puebloan (Gumerman & Gell-Mann 1994: 19). This was termed Basketmaker III, and included new architectural features such as Great Kivas, manifestations of inter- or intra-community ritual interaction (Gumerman & Gell-Mann 1994: 19). These Great Kivas developed architecturally from the pithouse, and similarly were semisubterranean structures (Fagan 2000: 320). Another architectural development was the construction of adjacent, above-ground dwellings and storage rooms, resulting in the creation of pueblos, the archetypal form for which the people were named (Gumerman & Gell-Mann 1994: 19). These architectural forms were standardized across the Ancestral Puebloan culture area, indicating a degree of social connectivity throughout the area. However, this does not indicate any degree of social control throughout the area (Gumerman & Gell-Mann 1994: 20). D. Pueblo I (AD 750 – AD 900) The developments of the Basketmaker III period continued into the Pueblo I period, and there is some difficulty distinguishing between the two. Previously, archaeologists separated Basketmaker III and Pueblo I on the basis of the presence of pueblos and the absence of pithouses, but it is now apparent that both architectural forms were used simultaneously in the same settlements, so the distinction between the two is false. The recognizable features of the Pueblo I period are a continued growth of population and a growth of complexity in agricultural and irrigation projects (Cordell 1997, 280; Fagan 2000: 318). Architecturally, communities continued to use traditional forms such as pithouses, but above ground pueblos were gradually 14 replacing the pithouses (Fagan 2000: 317). Unlike before, Pueblo I communities were perennial establishments, dependent upon agriculture to feed the population year round (Fagan 2000: 310). E. Pueblo II (AD 900 – AD 1150) Beginning circa AD 900, the Pueblo II period witnessed an explosion of population growth through the Ancestral Puebloan area. By AD 1000, the Ancestral Puebloans had reached their greatest territorial extent, occupying ―virtually every conceivable spot‖ (Cordell 1997: 280). This coincided with a considerable increase in social complexity, with evidence of regional hierarchy and region-wide cultural systems and trade patterns, namely that of the Chacoan system, centered in Chaco Canyon, New Mexico (Gumerman & Gell-Mann 1994: 20-21). This complexity even spun off smaller, daughter cultures such as the Kayenta, Virgin, and Winslow branches of the Ancestral Puebloans (Cordell 1997: 280). The Chacoan system evolved from a need to manage the scarce natural resources of the San Juan Basin and is now evidenced by sophisticated small-scale irrigation works, especially in Chaco Canyon, and the planning of large pueblos, both within Chaco Canyon, and 70 satellite communities named ―outliers‖ scattered throughout northwest New Mexico and southwest Colorado (Fagan 2000: 324). The nature and degree of the Chaco system, however, is a subject of considerable debate within American archaeology, but the fact that the Chacoan system was a substantially new phenomenon is not debated. As Gumerman & Gell-Mann state: ―The geographic extent of the large systems and of their exchange networks—and in the case of Chaco the obvious central planning of large towns—indicates a level of information sharing that required a high degree of coordination. The extensive regional systems were not simply aggregations of smaller local traditions but were qualitatively different in organization‖ (1994: 21). The last century and a half of the Pueblo II period saw the contemporaneous expression of many different cultural forms, but this in itself did not increase the amount of variation 15 expressed in the region. For example, what was once the Mogollon culture in the highlands of the central Southwest lost its distinctiveness and began expressing Ancestral Puebloan traits (Gumerman & Gell-Mann 1994: 22). At the end of this period, a severe drought developed in AD 1130 throughout the Colorado Plateau and despite the expansion of population and culture throughout the region, several cultures experienced decline or even collapse (Fagan 2000: 328). F. Pueblo III (AD 1150 – AD 1350) The start of the Pueblo III period circa AD 1150 saw striking population movement and sociocultural realignment throughout the Southwest (Cordell 1997: 365-397). Within the Ancestral Puebloan area, the Virgin Branch area in southern Nevada and Utah experienced a population collapse and the Chacoan system as a whole experienced a political collapse (Gumerman & Gell-Mann 1994: 22). Throughout the Ancestral Puebloan region, sites were more clustered, indicating territorial reduction and in some places a decrease in organization complexity (Gumerman & Gell-Mann 1994: 22). However, even after the collapse of the Chacoan system, Chaco Canyon was not completely abandoned and some construction continued, although in architectural forms apparently imported from the Mesa Verde to the north. Chacoan outliers, especially those located in the Middle San Juan Region, continued to function as community centers, but exhibited Mesa Verdean influence rather than Chacoan (Fagan 2000: 329). The Mesa Verde and Sinagua (located near Flagstaff, Arizona) regions experienced a growth in population, likely due to migration from other areas of the Southwest. This population growth was accompanied by an increase in construction efforts, and there is artifactual evidence for an increase in interaction with other cultural areas (Gumerman & GellMann 1994: 23). 16 These changes were likely a result of severe environmental stresses introduced by the Great Drought that occurred during the final quarter of the Twelfth Century AD. However, by AD 1250 much of the population growth experienced by larger sites was reversed, and there was again a period of large scale regional abandonment, resulting with the collapse of sites throughout Mesa Verde, the Middle San Juan region, and Sinagua regions which were thoroughly emptied (Fagan 2000: 338). Much of the Ancestral Puebloan population relocated to areas now occupied by the modern Puebloan groups, and some even left the Southwest, moving to sites such as Casas Grandes in Chihuahua and Trincheras in Sonora (Gumerman & Gell-Mann 1994: 23). G. Pueblo IV (AD 1350 – AD 1600) The Pueblo IV period was a time of massive realignment after the collapse of the Chacoan system. Population movement depleted formerly densely populated areas of the Southwest. As stated before, areas which enjoyed a greater supply of summer rains and perennial rivers, such as the Phoenix Basin, the Rio Grande Valley, the Zuni region of west-central New Mexico, and neighboring areas to the south absorbed the population (Fagan 2000: 343). Simultaneously, populations aggregated from smaller villages to large communities, some of which were larger than the largest pueblos of earlier times (Fagan 2000: 344). It was the establishment of these larger settlements that led to the development of the modern Puebloan cultures, and some of these settlements lasted for hundreds of years (the Hopi pueblo of Oraibi, Arizona was founded sometime before AD 1100 and is the longest continuously occupied settlement in the United States) despite impending contact with Europeans and the introduction of foreign pathogens and technology (Fagan 2000: 345). 17 V. Historical Review of Remote Sensing Techniques The history of remote sensing for archaeological purposes extends several centuries, and the principles of remote sensing have been correctly understood just as long. Because of the time-sensitive nature of archaeology, remote sensing techniques, which promise the delivery of large amounts of data in little time, have been adapted from other disciplines to great success. This has continued with the homegrown development of other analytical techniques, such as phosphates analysis, for use by archaeologists in their own research. This section will trace the development and evolution of remote sensing techniques from their beginnings to their modern forms and applications. However, it will focus primarily on aerial and satellite photography, as it is those techniques that have been used most successfully for archaeological surveying and because those techniques are employed in this study. The four geophysical techniques (electric resistivity surveying, electromagnetic surveying, magnetometry surveying, and ground penetrating radar surveying) and analytical chemistry techniques will each be discussed briefly. A. Aerial Photography In the sixteenth century, the English antiquarian William Camden discussed the occurrence of cropmarks in grain fields. Cropmarks are patterns of differential growth resulting from underlying conditions of the soil, and Camden correctly reasoned that the intersecting patterns of low crop growth (colloquially known as St. Augustine‘s Crosses) found in fields throughout England were the result of the presence of ancient roads within Roman sites buried underneath (Parrington 1983: 105). These former roads negatively influenced the growth of plants, and the resulting patterns reflected the grid nature of Roman era streets. Though his interpretation was common knowledge among antiquarians of that age, the knowledge of the 18 relationship between underlying archaeological features and overlying crop growth was not put to practical use until the early 18th century. At that time, another antiquarian, William Stukely, created the first archaeological plan of a cropmark, which he observed through a second story window of an alehouse (Parrington 1983: 105). This cropmark was formed from the presence of the buried foundations of a Roman era temple, and Stukely was able to accurately map the temple foundations without actual excavation. At this point, it was understood that cropmarks could convey much information concerning archaeological sites, but the utility of mapping cropmarks was limited to those areas where a difference of elevation could allow access to an overhead view of the cropmarks. During World War I, it was quickly realized that the newly invented airplane could be used as a platform to observe these cropmarks. Archaeological sites were noticed by pilots during the war, but it was during the post-war period that aerial photography of archaeological sites was first accomplished in large scale, as pilots who overflew the sites during the war returned to satiate curiosity (Parcek 2009: 14). Pioneering this development was the English archaeologist O. G. S. Crawford, who overflew much of southern England after the war (Parcek 2009: 15; Parrington 1983: 107). He and his partner, Alexander Keiller, eventually published the illustrative and interpretive work Wessex from the Air in 1928. Likewise, in 1929 Charles Lindbergh overflew and catalogued Pueblo sites in the American Southwest and Mayan sites throughout Central America (Deuel 1973). As a result of these early efforts, aerial photography was recognized as a valuable tool for archaeological prospecting and has been used throughout western Europe and the United States (Parrington 1983: 108). Aerial photography and its associated techniques have developed over the past ninety years. Black and white photography revealed the same crop mark features that were visible from 19 high windows, while color photography could reveal slight differences in the color tone of plants which could indicate the presence of subsurface features affecting the growth of plants, but both of these were limited to recording the phenomena that were visible to humans (because human beings had to first observe the phenomena before photographing it, and also because of the technical limitations of the photographic film at the time). However, the development and widespread availability of infrared photography in the 1930s allowed for the discovery of invisible soil patterns and the differential release of residual heat due to alterations in the soil and bedrock (Berlin et al. 1990). Despite the utility of aerial photography, only four classes of archaeological features can be recorded by such techniques: cropmarks, soilmarks, shadowmarks, and snowmarks (Parrington 1983: 108, Scollar et al. 1990: 33-58). The first two are relatively permanent features of the soil, whereas the last two are temporary phenomena that can be detecting and subsequently used to locate buried archaeological features. As discussed previously, cropmarks are areas of differential growth created by underlying archaeological features. The marks may be either a decreased growth rate or an increased growth rate: better growth is the result of plants growing in an area where the archaeological artifacts and features contribute nutrients or trap moisture to aid plant growth. Conversely, decreased growth is a result of plants being located over features or artifacts that stunt growth, such as stone foundations and walls that limit root growth and therefore limit plant development (Scollar et al. 1990: 50-52). Soilmarks are areas of differing soil color; these differences can be either in the visible spectrum or in portions of the spectrum outside the human visible range. They can be observed in plowed fields before plant growth obscures such differences, or in other areas of bare soil such as deserts, where differences in soil color can be detected by a number of above-ground surveying techniques 20 (Altaweel 2005). Shadowmarks are temporary patterns detectable only in the early morning and late evening, when the sun is low to the horizon. The position of the sun low on the horizon results in exaggerated shadows. Low earthworks, which normally would not be visible to the eye, might cast these shadows that highlight their location (Parrington 1983: 108). Finally, snowmarks are the result of deferential melting of light snow, whether it is caused by differences in heat retention by the underlying soil features or by a process similar to shadowmarks where the snow is protected from the sun‘s rays by slight changes in elevation due to underlying features (Scollar et al. 1990: 49). The presence of snow might highlight these slight changes in elevation for short periods, and while positive results have been achieved from the use of this method, it is by definition ephemeral and lack a consistent correlation to underlying archaeological features (Parrington 1983: 108). B. Electric Resistivity & Electromagnetic Surveying Aerial photography remained the only form of remote sensing available for archaeology until 1946 when resistivity surveying was developed by Atkinson (Scollar et al. 1990: 307, 371). This relatively simple and inexpensive method measures the electrical conductivity of the soil, hopefully illustrating the presence of non-conductive elements which could be archaeological features. The technique involves the deployment of a pair of electrodes into the ground, and an electrical charge is passed from one electrode to the other through the soil. The resistance of the earth to the electrical charge is measured, and differences in resistance indicate differences in the soil, which may or may not correspond to archaeological features. Parrington (1983) states that the method works best in damp soil, and is most successful in locating buried stone walls (as the electrical charge can simply bypass smaller features). Scollar et al. (1990: 359-371) details, in depth, the effects of various features on the electrical resistivity of the soil. Due to the 21 nature of electricity, resistivity surveying cannot be used in dry soils—there simply is no way for the electrical charge to be carried from one electrode to the other. Resistivity surveying showed initial success, but the first application of the method was conducted with ideal soil conditions on a site outside of Dorchester, England, a feat not easily replicated in subsequent studies and the method was shown to be limited (Parrington 1983: 106). There are other disadvantages: the time and effort involved to move and set the probes, as well as recording the resistivity, made the use of electrical resistivity surveying for large areas unrealistic (Scollar et al. 1990: 335). As a result, other related methods were developed, such as electromagnetic surveying (Parrington 1983: 114). This method also measures the ground‘s resistance to electricity, but unlike electrical resistivity surveying, the EM surveying instrument does not need to be placed in the soil and so can be moved from location to location quickly. It is usually mounted on a wheeled cart, allowing it to be pushed or pulled along predetermined paths called ―instrument tracks.‖ It provides a continuous readout of the electrical resistivity along the instrument track, allowing for quicker implementation of recovered data (Aitken 1974: 191-198, Parrington 1983: 114). The development of electromagnetic meters and their associated surveying techniques has advanced this particular field of remote sensing (Parrington 1983: 114). It can be used by a single person, another improvement on electric resistivity surveying, which required groups of people to set and move the electrodes. Due to the nature of electromagnetic surveying, EM meters can deliver verifiable results in dry soils, something that was not possible with electric resistivity surveying. Finally, EM meters can detect ditches or the remains of earthen mounds; these features are differences in soil rather than differences of material such as the stone walls detected by electric resistivity surveying (Parrington 1983: 114). 22 C. Magnetometry Related to electromagnetic surveying is magnetometry. Like EM surveying instruments, the magnetometer is deployed on a wheeled cart and pushed along tracks, and measures slight disturbances in the magnetic field of the earth, mapping out the location of such disturbances. These disturbances are the result of culturally derived disturbances in the soil, specifically the presence of iron artifacts or burnt features in which the temperature of the feature has been raised above the Curie point, the point at which magnetic grains are ―reset‖ and the direction of the grains become random in relation to the magnetic field of the earth (a common occurrence in hearths and furnaces). The resulting randomness of the magnetic grain is easily detected by the magnetometer. Pits and linear features, such as ditches, are quickly determined, but the instrument can also detect walls and stone structures that have not reached their Curie points, as many stone materials contain varying amounts of ferrous compounds that, unless perfectly aligned with the Earth‘s magnetic field, can also be detected by the magnetometer (Parrington 1983: 115; Scollar et al. 1990: 435-438). After a four year period of theoretical and methodological development, magnetometry was first deployed as a prospecting technique in 1958 (Scollar et al. 1990: 514), and has exhibited much success in archaeology: Scollar et al. (1990: 450) states ―it is probably not an exaggeration to state that the … magnetometer has been more widely used in archaeological prospecting than all other instruments taken together.‖ The technique‘s utility is notable, especially when it is used to map out the extent of individual sites prior to excavation, such as at the site of Titrus Hoyuk, Turkey (Matney & Algaze 1995; Parrington 1983: 106). Magnetometers work best in large, open areas with no magnetic disturbance from ferric objects such as vehicles and utility objects like power lines and sewer pipes (Parrington 1983: 115). 23 However, the technique is often limited to the uppermost occupation level, yet if one wants to map out the most recent site features, it is a useful and inexpensive alternative to traditional excavations (Matney & Algaze 1995: 299). D. Ground Penetrating Radar & Sonar Following the development of magnetometry is the use of ground radar, which has proven to be one of the most useful of all remote sensing techniques (Parrington 1983: 115). Developed in the early 1970s and first deployed at Chaco Canyon National Historical Park, New Mexico in 1974, it uses the same principles as the aerial radar used by aircraft and weather radar deployed throughout the world (Scollar et al 1990: 575-580). Like EM surveying and magnetometry, a radar antenna is deployed on a wheeled cart where it can transmit signals into the ground. If the radar waves hit a disturbance, the waves are then bounced back to the radar antenna, which records and displays the data on a monitor. Ground radar can detect small and weak variations of soil density, except when used in clayey soils (Scollar et al 1990: 582). Under the right circumstances, such as dry volcanic soils, it has detected stratigraphic layers of archaeological sites, and features and artifacts of various sizes almost as well as traditional excavation techniques (Parrington 1983: 117, 119, Scollar et al 1990: 584). Ground radar does not suffer the same interference that affects magnetometry; however, limitations of the technique include the inability to determine the identity and composition of the targeted features, their depth, and their volume (Scollar et al 1990: 582-583). In other words, it can indicate that something is there, but nothing else. Also, the relative high price of the equipment (even to this day) and the inability to use it in thickly wooded areas or on rough terrain, as trees block the movement of the radar over the ground (Parrington 1983: 115). 24 Similarly, sonar, also known as acoustic sensing, was developed in Japan and has been used in much the same manner (Parrington 1983: 120). A sound generator sends a lowfrequency acoustic pulse into the earth, which is then reflected by artifacts and features. This pulse is then received by the antenna and the signals are interpreted and displayed on a monitor, much like ground radar, and the utility of the tools are similar. E. Analytical Chemistry Techniques Though they are not technically remote sensing, analytical chemistry techniques have been used to great success by many archaeologists throughout both Europe and North America, with successful applications also in Mesoamerica and the Middle East. The most widely used technique is phosphate analysis, which involves the removal of soil samples from the field to the lab for chemical processing to determine relative or absolute concentrations of phosphates, chemical compounds derived from human activity (see Holliday & Gartner 2007 for an exhaustive list of phosphate analyses in archaeology). Phosphate analysis in archaeology was first developed in the early twentieth century, and like many archaeological techniques and principles it first developed elsewhere, in agricultural studies (Bethell & Máté 1989). Russell (1957) claims that Hughes in Egypt in 1911 first noticed that human occupation increases the concentration of phosphorus in soil over time, and that this concentration will stand out over a background phosphorus content. However, it was not until the early 1930s that this property was used in an archaeological context. Olaf Arrhenius, a Swedish agronomist, also noticed that raised phosphate levels occurred in conjunction with human settlement while he worked with the Swedish government to map soil conditions (Bethell & Máté 1989). However, Arrhenius developed this principle further and realized that one could 25 locate settlements by measuring for enhanced phosphate levels (Bethell & Máté 1989). He then used this principle in the first large scale systematic archaeological analysis of Sweden. In the 1960s, the archaeological discipline was focusing on finding more scientific methods to conduct archaeology, and phosphate analysis was more widely used as a surveying technique. In 1965 Cook and Heizer authored a paper entitled ―Studies on the Chemical Analysis of Archaeological Sites.‖ In this paper, the authors argued that an understanding of the entire soil system is necessary in order to interpret any recovered data concerning soil concentrations, and that phosphates in particular cannot be considered without also considering the entire environment (Cook & Heizer 1965). Cook and Heizer‘s advice was followed, and during the 1970s phosphate analysis was taken in conjunction with other remote surveying methods, such as resistivity and magnetometry surveying and aerial photography (Bethell & Máté 1989). Also, phosphate analysis was being used on a much smaller scale, this time in an inconclusive attempt to determine whether or not a body was at one time present in the Sutton Hoo ship (Barker et al. 1975). The use of phosphate analysis continued to evolve during the late 1970s and 1980s with the advent of processualism, a movement toward scientifically quantitative archaeology. In 1973, Provan excavated a Norwegian farmstead, and in his report included the results of phosphate analysis as well as soil descriptions, density, and elemental measurements (Provan 1973). This follows with his statement that phosphate analysis was able to locate a site, determine its limits, illustrate the diet of its inhabitants, and determine the use of structures (Provan 1971). At Cat‘s Water, in Fengate, Great Britain, phosphate analysis confirmed the existence of a deeply stratified site prior to excavation in 1984. Also, the phosphate analysis was used to map the three-dimensional extant of the site prior to the excavation. A second round of testing was 26 including during the excavation to aid in the interpretation. Finally, the use of phosphate analysis prior to excavation illustrates a step towards adopting phosphate surveying as a standard technique (Bethell & Máté 1989). In 2000, E. Christian Wells coauthored a paper describing the chemical analysis of anthrosols at Piedras Negras. He first states that ―a number of phosphate compounds are associated with human activities, such as food preparation and consumption, and disposal of food products and fecal materials‖ (Wells et al. 2000). On top of this, he illustrates that phosphate analysis can differentiate between elite compounds and structures associated with non-elites (Wells et al. 2000). Other chemical analyses can determine if structures were painted or the contents of middens (Wells et al. 2000). These analyses indicate the methodology bridges the role of remote sensing methods to a more traditional, archaeological analysis of past human activities. F. Satellite Photography During the 1960s, various governments began launching satellites with the capability to observe features of the earth from their vantage point in orbit. Naturally, the first satellites served the interests of the government bodies funding the development and deployment of the technology, and were primarily used for reconnaissance of enemies or potential enemies, such as the Soviet Union and its dependencies. While the instruments of these satellites were usually focused on sites of military importance, many times they observed archaeological and historical sites. An example of this was the observation of the historic city of al-Raqqa, Syria, by the Corona program. Because these images were classified until the 1995, they remain underutilized, but they represent the first large body of remotely sensed images (and as Challis 27 et al. 2002 illustrates, they offer a cost-friendly alternative to other, though more recent, photographs). The military and civilian development of remote sensing satellites was a continuation of the development and utilization of aerial photography, and early satellites were limited to the same portions of the electromagnetic spectrum that photography was: the visible and near infrared portions of the spectrum expressed through black and white or color photographs. The first satellite system to take pictures of the Earth and its atmosphere was the Television and Infrared Observation System (TIROS), a meteorological observation system launched by the United States in 1960 (Parcek 2009: 19). The United States military quickly saw the utility in a space-based observation system, and launched a series of military reconnaissance satellites, with specific satellite systems codenamed Corona, Argon, and Lanyard. These satellites, eventually numbering six in total, were designated the ―Keyhole‖ program and took images of the surface of the earth from 1960 to 1972 (Parcek 2009: 19). During the 1960s and 1970s, methodological development of archaeological uses of aerial photography demonstrated that the observable portions of the EM spectrum needed to be expanded from visible and near infrared to incorporate other, previously unused portions such as the microwave and far infrared portions (Parcek 2009: 20). This methodological development, along with coinciding theoretical and problem-orientated developments in the new subfield of archaeological remote sensing, was applied to satellite systems as well as airborne sensors. During the late 1960s and early 1970s, The U.S. Department of the Interior developed the Earth Resources Technology Satellite (ERTS), with the goal of achieving a non-militarized (and non-supervised) satellite observation system with multispectral capabilities for peaceful purposes. The first satellite of the program was launched in 1972, and the United States 28 government invited scientists from throughout the world to study and analyze the data obtained from ERTS-1 (Parcek 2009: 22). In 1975, the program and its seven associated satellites were renamed ―Landsat.‖ The Landsat program provided the first space-borne data applicable to archaeological investigations on the earth, with most of the data related to the geomorphology of large geographic regions (Parcek 2009: 22). Likewise, in 1978, the government of France began development of another multispectral satellite system named the Systeme Pour l‘Observation de Terre (SPOT). Launched in 1986, SPOT was the first satellite system capable of stereoscopic imaging (Parcek 2009: 22). Early efforts to incorporate satellite imagery for archaeological purposes proved quite successful, and several projects investigating prehistoric and anthropogenic landscapes throughout northern Africa and the Yucatan were able to effectively employ the Landsat system despite an 80 meter spatial resolution (Parcek 2009: 26). These successes led to the organization of the first conference on Remote Sensing in Archaeology in 1984, and the papers and discussions from this conference resulted in the continued application of Landsat and other satellite imagery to regions throughout the world: archaeological investigations in England (Donoghue & Shennan 1987, 1988), Sweden (Lunden 1985), Delaware (Custer et al. 1986), Syria and Palestine (Cleave 1985), Libya (Dorsett et al. 1984), and Egypt (Wendorf et al. 1987; McHugh et al. 1988, 1989) resulted in seminal works on space-borne remote sensing applications for archaeology. Applications of satellite imagery for archaeological projects increased throughout the 1990s, and the topic of the integration of remote sensing imagery with Geographic Information Systems (GIS) technology was first addressed (Parcek 2009: 28). GIS was able to easily integrate knowledge of previously determined areas of environmental and cultural importance 29 with areas under investigation, but due to a lack of training GIS remained relatively unused by archaeologists and this fact was critiqued by archaeologists (Parcek 2009: 29). Also, while there was an emphasis on the development of predictive models in other fields, there was a noticeable lack of articles on the topic and theoretical ignorance of the issues of predictive modeling was not addressed within archaeology (Davis 1993). Finally, the use of satellite imagery in archaeology was discovered by global media, and successful projects from New Mexico and Costa Rica (Sheets and Sever 1991), Greece (Joyce et al. 1992), Thailand (Parry 1992), and especially the discovery of the lost city of Ubar in Arabia (also known as Iram, or the City of a Thousand Pillars, and mentioned in the Qur‘an) were brought to the public‘s eye (Fiennes 1991; Blom 1992). Satellite technology continued to develop during this same time. Microwave and radar sensors were deployed on space-borne systems, and the spatial resolution of the satellites increased. SPOT 4 was launched in 1998 with a spatial resolution of 5 meters, and IKONOS was launched by Lockheed Martin in 1999 with panchromatic and multispectral band spatial resolution of 1 meter and 4 meters, respectively. Quickbird, with a panchromatic band spatial resolution of 0.6 meters and a multispectral band spatial resolution of 2.4 to 2.8 meters was developed and launched by the private corporation DigitalGlobe in 2001. Finally, hyperspectral sensors, instruments that can observe the EM band in hundreds of individualized portions, were developed and deployed on a variety of satellite vehicles. These improvements in sensor sensitivity continue to aid archaeological investigations, as the increased resolution allowed the discrete discernment of increasingly smaller portions of project areas. 30 G. Application of Remote Sensing Techniques to the Archaeology of the Southwest Despite the growing number of archaeological projects employing remote sensing, there are very few publications on the topic of the utilization of space-borne remote sensors to determine landscape patterns within the American Southwest. Past projects have included nearground surveying techniques such as magnetometry and ground radar (Scollar et al. 1990: 575580), and there are examples of airborne surveying projects (Deuel 1973 discusses Charles Limburg‘s flights over Chaco Canyon, New Mexico) or projects where airborne sensors were used to track the condition of archaeological sites (Palumbo & Powlesland 1996). Perhaps the best known project to incorporate the use of aerial imagery may be George Gumerman and Thomas Lyons‘ 1971 study ―Archaeological Methodology and Remote Sensing‖ which was the first overview of remote sensing in an archaeological publication. In it, Gumerman & Lyons discussed the utility of aerial multispectral remote sensing and demonstrated the ability to differentiate between soils, rock types, vegetation, geomorphology, and sites. They were also able to perform soil moisture analysis, vegetation pattern analysis, plant vigor analysis, and land use analysis (Gumerman & Lyons 1971). Despite the clear utility demonstrated by the article, it appears to have been unheeded by Southwestern archaeologists. As mentioned previously, the first conference on Remote Sensing in Archaeology was held in 1984, and discussions concerning the remote sensing of Chaco Canyon were included. One of these, presented by conference organizer Tom Sever, discussed the inability of Landsat to aid most archaeological investigations due to the instruments‘ 30 meter spatial resolution (Parcek 2009: 26); Sever switched to an airborne Thermal Infrared Multispectral Scanner to continue his project (Sever & Wagner 1991). Though the poor spatial resolution of past instruments is admittedly a factor, Sever‘s articles may have negatively impacted the field, as very little 31 research has been done on the application of space-borne remote sensing despite the increase spatial resolution of modern sensor technology. It is the goal of this project to apply such technology to an area in which it has not yet been applied. 32 VI. Literature Review This chapter will review three different approaches to the application of remote sensing techniques to archaeological problems. They involve various methods to classify multispectral imagery obtained from two space-borne instruments. These instruments are the declassified Corona system, the IKONOS system, and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument onboard the Terra satellite launched by NASA in 1999. Unlike Corona, which delivered black and white images of the visible spectrum, ASTER is a multispectral instrument capable of delivering true and false color images from the visible, near infrared, short-wave infrared, and long-wave infrared/thermal portions of the electromagnetic spectrum. ASTER data has traditionally been used for geologic, environmental, and population studies (Altaweel 2005: 151), but because of the capabilities of the ASTER instrument, ASTER data was chosen for this project. Because the approaches reviewed by this chapter have successfully implemented multispectral data within the semi-arid environment of the Middle East (specifically portions of the nations of Syria, Turkey, and Iraq), they have been selected to determine if any of the approaches may be successfully implemented in the Middle San Juan Region of New Mexico. A. “The Use of ASTER Satellite Imagery in Archaeological Contexts” This article, written by Mark Altaweel and published in the journal Archaeological Prospection in 2005, investigates the applicability of ASTER data for determining large-scale archaeological features in northern Mesopotamia, in northern Iraq. These features include roads (called ―hollow ways‖ by Altaweel), canals, and sites that could not be easily distinguished within Corona images due to the inability of the Corona system to detect separate spectral signatures for portions of the electromagnetic band outside the visible range, despite the two to 33 three meter resolution of the Corona cameras (Altaweel 2005: 151). By using ASTER data, the author was able to verify the presence of features located in the Corona imagery, as well as discover features not previously discernable within Corona images. The first feature class investigated by Altaweel, road beds (or hollow ways), were investigated using a simple qualitative technique. After performing a series of conversions and analyses on bands 1 through 9 of the ASTER image in question (corresponding to the visible, near infrared, and short-wave infrared portions of the EM spectrum, which can indicate the presence of water, plant health, and different soils and minerals), Altaweel visually compared known roadbeds with similar features. The known roadbeds and the unknown features were qualitatively similar, and exhibited similar patterns such as direction and terminating at known archaeological sites (Altaweel 2005: 157). As such, he was able to successfully identify previously unknown roadbeds. Canal features were similarly investigated. However, these features were verified using the spectral analysis capabilities of the ASTER instrument, which permitted the comparison of known feature spectral signatures with the spectral signatures of potential features (Altaweel 2005: 158). Rather than being a purely qualitative comparison between known and unknown but similar features, this method allowed for a quantitative comparison of known and suspected canalways. However, Altaweel admits that similar spectral signatures of two features only indicates a degree of likelihood that the two features are of the same type, and that field work needs to be undertaken to positively identify the suspected feature as a canal (Altaweel 2005: 160). Finally, Altaweel investigated potential archaeological sites throughout the ASTER imagery by employing a supervised classification method. Supervised classification methods 34 involve the identification of known sites in an image. These known sites then are analyzed, and a range of reflectance values for a subset of the spectral bands are identified. Suspected sites are then compared to this range of values, known as a classifier, to determine if they exhibit similar reflectance values. This method is also a quantitative method, and Altaweel was able to identify a previously unknown archaeological site in the image (2005: 161). Again, the author stated that a field visit was required to determine positively if the suspected site was truly an archaeological site. B. “Classification of Multispectral ASTER Imagery in Archaeological Settlement Survey in the Near East” This paper, authored by Bjoern H. Menze and Jason A. Ur, was presented at the Tenth International Symposium on Physical Measurements and Signatures in Remote Sensing, a symposium organized by the International Society of Photogrammetry and Remote Sensing in 2007. It is available in the Proceedings of that same symposium. Similar in methodology to Altaweel (2005), it discusses a project to identify archaeological sites in the Middle East by developing a spectrum signature. However, the authors go further than Altaweel by illustrating how differences of climate, time of day, angle of the sun, and season can mask or alter the spectrum signatures of archaeological sites (Menze & Ur 2007: 1). To counteract these effects, the authors combine all available ASTER images of their study area (16 total) and average the spectrum signatures to obtain a composite signature (Menze & Ur 2007: 3). By using the composite signature in a supervised classification method, the authors were able to obtain a correct classification rate of approximately 80%, significantly higher than the average of 60% correctly classified when they used spectrum signatures obtained from individual ASTER images (Menze & Ur 2007: 3). By combining multiple ASTER images, 35 the authors were able to develop a methodology that consistently outperformed methodologies reliant upon single observations. It also relieves the investigator of the choice of selecting the ―best‖ available image, because all images of sufficient quality are used to develop the classifier. However, it is often difficult to obtain multiple ASTER images of sufficient quality to duplicate this technique. Finally, the authors discuss the importance of understand false-positive and falsenegative results, and discuss the potential of using a classifier explicitly trained to identify problematic results, such as archaeological sites located near modern settlements (Menze & Ur 2007: 6). They claim that ―false positives in the area of modern settlements might result from covariates in the (current) training data, or might resemble the spectral characteristic of ancient debris indeed, eventually indicating the presence of (unrecorded) former sites at the same place‖ (Menze & Ur 2007: 6). These issues need to be kept in mind when developing a supervised classification scheme, otherwise sites may go undetected. C. “Archaeological Site Detection: The Importance of Contrast” This paper, written by A. Beck and delivered to the Annual Convention of the Remote Sensing and Photogrammetry Society in 2007, also focuses on archaeological sites in the Middle East. However, rather than relying solely on a supervised classification method that depends upon the spectral signatures of archaeological sites differing from the signatures of non-site areas, the author developed a method that relied upon the relative amounts of reflectance detected by the IKONOS instrument. Beck states: ―Unfortunately archaeological sites do no exhibit spectral signatures that can be used for generic detection purposes (however, see Altaweel (2005) for use of spectral signatures in a constrained environment). Rather, it is hypothesized that archaeological residues produce localised [sic] contrasts in the landscape matrix which can be detected using an appropriate sensor under appropriate conditions‖ (Beck 2007: 1). 36 While we know, and Beck even admits, that his claim is not entirely true, Beck does develop a unique methodology to detect archaeological sites. He compares his methodology to the detection of soilmarks in aerial photography and Corona imagery, noting that archaeological residues, when dry, were significantly lighter in color than the surrounding soil matrix (Beck 2007: 4). Beck claims that the contrast is not only visual, but would also extend outside of the visual wavelengths. Beck‘s methodology attempts to detect sites not by difference in the spectral reflectance, but rather by the strength of reflectance. He states that many of the enhancements undertaken to enhance the visual contrast of archaeological sites to their surrounding matrix mask the subtle reflectance differences that may be more effective in determine the presence of archaeological remains (Beck 2007: 5). Rather, the presence of any differences in the strength of signal reflectance can be highlighted by the application of a moving average kernel based upon the average background soil pixel. After the application of the filter, any areas of unmodified soil should have an average approaching zero, but areas significantly deviating from the background values (such as archaeological sites and modern construction) should exhibit positive or negative values (Beck 2007: 5). D. Discussion These papers illustrate increasing levels of sophistication in the interpretation of multispectral imagery, and the methods presented in the papers were analyzed for this project. All three papers discuss the utility of supervised classification methods, though the authors might disagree on its use. While Altaweel‘s methods are relatively simple compared to the other authors‘, his use of supervised classification is common, and is repeated in this project. Likewise, Menze & Ur employed supervised classification methods, but mentioned the importance of 37 minimizing false-positive and false-negative errors. As such, this project takes into account the methods discussed by Menze & Ur to avoid, as much as possible, those errors. Unfortunately, due to the limited resources of this project, it will not attempt Menze & Ur‘s method to combine and average the spectral signatures of multiple ASTER scenes. Finally, Beck, who disagreed with the use of classification methods, created a method to determine the presence of archaeological sites by highlighting reflective differences that is simple, but unique. Beck illustrated that many of the traditional analyses actually mask subtle differences that may indicate the presence of archaeological sites; this must be accounted for when designing a project. 38 VII. Methodology This chapter will discuss the ASTER instrument and the images taken by that instrument, and also the methodology undertaken to analyze the ASTER imagery for the project. The instrument was launched in December 1999 onboard the satellite Terra, the first craft launched in the multi-craft series ―Earth Observing System‖. It was given the designation EO-1 as a result. Terra and the Earth Observing System is a multinational cooperative mission, with teams of scientists located in the United States, Canada, and Japan. The satellite‘s primary mission is to ―research into the ways Earth‘s land, oceans, air, ice and life function as a total environmental system‖ (JPL 2009). It orbits the Earth in a sun-synchronous orbit at 705 kilometers, and is inclined 98.3 degrees from the equator. Its orbit period is 98.99 minutes, crossing the equator at 10:30 AM. It takes 16 days for the ground track to repeat itself; this is known as its temporal resolution. The vehicle carries five instruments: ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer), CERES (Clouds and the Earth‘s Radiant Energy System), MISR (Multi-angle Imaging SpectroRadiometer), MODIS (Moderate Resolution Imaging Spectroradiometer), and MOPITT (Measurements of Pollution in the Troposphere) (JPL 2009). ASTER‘s primary missions are related to the environment: it was launched to monitor land surface climatology, vegetation and ecosystem dynamics, volcano and other hazard monitoring, hydrology, geomorphology, and land cover change (JPL 2009). However, as mentioned before, it has found some utility in archaeological projects, particularly the surveying of regions for settlement sites. It conducts these missions with three instrument subsystems, which focus on different portions of the electromagnetic spectrum. These are listed in Table 2, adapted from Altaweel (2005). 39 Table 2: Subsystems of the ASTER Instrument, from Altaweel (2005) Subsystem Type ASTER Band Band width (µm) Spatial Resolution (m) VNIR Band 1 0.520 – 0.600 15 VNIR Band 2 0.630 – 0.690 15 VNIR Band 3n 0.760 – 0.860 15 VNIR Band 3b 0.760 – 0.860 15 SWIR Band 4 1.600 – 1.700 30 SWIR Band 5 2.145 – 2.185 30 SWIR Band 6 2.185 – 2.225 30 SWIR Band 7 2.235 – 2.285 30 SWIR Band 8 2.295 – 2.365 30 SWIR Band 9 2.360 – 2.430 30 TIR Band 10 8.125 – 8.475 90 TIR Band 11 8.475 – 8.825 90 TIR Band 12 8.925 – 9.275 90 TIR Band 13 10.250 – 10.950 90 TIR Band 14 10.950 – 11.650 90 Bands 1 and 2 focus on the red and green/yellow portions of the visible spectrum, respectively. Band 3 focuses on the near infrared, while Bands 4 through 9 focus on the shortwave portion of the EM spectrum, the portion sensitive to moisture, soil, and vegetation differences (Altaweel 2005: 153). As such, these bands were used by archaeologists to determine site location, and they were used in this project to attempt the same. 40 As stated before, this project was an attempt to determine archaeological site locations in the Middle San Juan Region of northwestern New Mexico. An ASTER image of the area was obtained on April 4, 2008, with the aid of Dr. Le Wang of the State University of New York at Buffalo. The image‘s upper left coordinate was centered above 108º 21‘ 10.85‖ West, 37º 5‘ 39.20‖ North. This translates to a Universal Transverse Mercator position of grid 12N, 735250.432 meters East, 4108603.669 meters North. The image has a rotation of -8.700343 degrees. This image was subset using ENVI 4.4 to include an area of 22.55 miles east-west by 19.66 miles north-south, with the upper left corner measured at 740859 meters East and 4090090 meters North. 41 Because the analyses used both VNIR and SWIR bands, the SWIR bands were resampled to match the 15 meter spatial resolution of the VNIR bands. This resampling was conducted in ENVI 4.4, and involved a first degree Rotation, Scaling, and Translation (RST) and nearest neighbor algorithms. A forward Principal Component Rotation using a covariance matrix was then conducted on both VNIR bands 1 through 3 and SWIR bands 4 through 9. Principal Component Analyses is used to produce uncorrelated bands; that is, the first of these bands contains the highest amount of variance, the second band contains the second highest, et cetera. This variance is the amount of data that is not directly related to its neighboring data. Three bands for each set were created; this correlates to an Eigenvalue of 14.222 for the VNIR band 3, and 0.0175 for SWIR band 3. Because VNIR only contains three bands, one hundred percent of information contained within the VNIR band was used to create the PCA VNIR bands, but because the SWIR information is initially held within six bands, only 99.87% of the information was used to create the three PCA SWIR bands. The resulting images then underwent a Hue Saturation Value (HSV) Sharpening color transformation. This transforms the image by resampling the hue and saturation values according Figure 4: Eigenvalue Graphs, Principal Component Analyses on VNIR and SWIR 42 to a nearest neighbor algorithm, but as it only masked the presence of previously located archaeological sites, the transformation was removed and succeeding analysis was conducted on the PCA bands. PCA VNIR bands 2 and 3, and PCA SWIR band 1 were then displayed in a Red-GreenBlue (RGB) colorspace. While multiple other combinations were tried, this combination presented the most data as determined by the principal component analyses while also maintaining a degree of visibility for the visual identification of previously recorded archaeological sites and modern features. It was upon this image that the classification techniques were attempted. While classification was attempted on other band combinations, this combination presented the most accurate classification. 43 The first set of classifications attempted was with the unsupervised IsoData classification technique included within ENVI 4.4. This technique calculates class means evenly distributed throughout the data space, and then iteratively clusters the remaining pixels using minimum distance (Tou & Gonzalez 1974). The software was instructed to create five to ten classes within twenty-five iterations on a variety of band combinations, including PCA VNIR bands 2 and 3, and PCA SWIR band 1. This was done to provide a visual reference for comparing future supervised classifications; however, while the classifications seem relatively accurate on a large scale, previously located archaeological sites were not consistently identified. This attempt to qualitatively match the classes of known archaeological sites provided a quick method to determine the accuracy of the various classification attempts. 44 After attempting a series of unsupervised classifications, supervised classifications of various kinds of undertaken. To conduct these classifications, a series of Regions-of-Interest (ROIs) had to be identified and selected. These regions corresponded to the various classes being identified: water, crops/plants, highland/bare rock, sand, road, urban areas, and archaeological sites. The ROIs for the classes followed the traditional ―ten times the number of bands‖ formula and were randomly selected from throughout the image, except for the case of the archaeological sites. The locations of the sites were determined previously using hand-held Garmin Global Positioning System (GPS) handsets. These locations were then brought into ENVI 4.4, and the ROIs corresponding to archaeological sites were developed. The shape of the individual site ROIs was determined from field visits and site records, and were attempted to match the physical site boundaries as much as possible. 45 It was decided that a mask would not be placed on the image to hide any one class (such as urban areas). The proximity of the various archaeological sites to the other classes, especially the urban class, was a factor in this decision. The fact that previous inhabitants of the region built in the same location as modern people seems unavoidable. As was discussed earlier, errors ―in the area of modern settlements might result from covariates in the (current) training data, or might resemble the spectral characteristic of ancient debris indeed, eventually indicating the presence of (unrecorded) former sites at the same place‖ (Menze & Ur 2007: 6). The use of a classifier explicitly trained to identify problematic results, such as archaeological sites located near modern settlements, might mitigate these issues (Menze & Ur 2007: 6). Three supervised classification methods were employed for this project. These were maximum likelihood, minimum distance, and spectral information divergence. Maximum likelihood assumes that the statistics for each class in each band are normally distributed and calculates the probability that a pixel belongs to a specific class; each pixel is assigned to the class to which it has the greatest possibility of belonging (Richards 1999: 240). On the other hand, minimum distance calculates the mean vectors of the classes and measures the Euclidean distance from each pixel to the mean vector of each class. It then classifies each pixel to the nearest class (Richards 1999: 240). Finally, spectral information divergence matches the spectrum of each pixel to a reference spectra obtained from the ROIs. The pixel is then matched to other pixels that exhibit similar spectra (Du et al. 2004: 1777 – 1786). These three methods were also performed on a variety of band combinations. However, one appeared to give the best result: the maximum likelihood method on the PCA bands of the VNIR portion. However, even this failed the qualitative assessment of whether or not the various archaeological sites were classified similarly. As can be seen in the image below, the 46 archaeological site class, colored red, has even been confused for crop growth, river sediment, and bare soil. Rivers (colored blue) and roads (colored magenta) are also confused, as are silt (cyan) and roads (magenta). Due to the obvious failure of the qualitative assessment, a quantitative measure such as a confusion matrix was deemed unnecessary. 47 VIII. Results & Conclusion The results of the analyses prove disappointing. The project was unable to develop and verify a methodology to detect archaeological sites in the project area. This may be due to the proximity of larger archaeological sites to modern urban areas; the spectral signature of those sites is masked by the larger urban areas. However, as Tom Sever mentioned in 1984, the spatial and spectral resolutions of the systems may not yet be at the point where they are able to distinguish between archaeological sites and their surrounding matrix (Parcek 2009: 26). This is true for ASTER imagery, whose 15 meter spatial resolution and 14 bands were unable to detect the smaller sites throughout the region, many of which are less than 15 meters square. Perhaps the implementation of other, more sensitive, sensors can advance the field of remote sensing for archaeological surveying in this area. The results of this project will be delivered to Linda Wheelburger, principal investigator for the Totah Archaeological Project based at San Juan College, Farmington, New Mexico, as well as delivered as a poster at the 2010 conference of Society for American Archaeology. It is desired that future work and collaboration of the project will bring about successes where the work for this project has only indicated the inadequacy of the methodology and sensors employed. 48 Works Cited Aitken, M. J. 1974. Physics and Archaeology. Oxford: Oxford University Press. Altaweel, Mark. 2005. The Use of ASTER Satellite Imagery in Archaeological Contexts. Archaeological Prospection 12:151-166. Barker, H. et al. 1975. Report on Phosphate Analysis Carried out in Connection with the Cenotaph Problem. In The Sutton Hoo Ship Burial Vol. 1 by Bruce-Mitford, R.L.S, pp. 550-72. London: British Museum Beck, A. 2007. Archaeological Site Detection: The Importance of Contrast. Paper presented at the Annual Convention of the Remote Sensing and Photogrammetry Society. Berlin, G. Lennis; David E. Salas, and Phil R. Geib. 1990. A Prehistoric Sinagua Agricultural Site in the Ashfall Zone of Sunset Crater, Arizona. Journal of Field Archaeology 17(1):1-16. Bethell, Philip & Ian Máté. 1989. The Use of Soil Phosphate Analysis in Archaeology: A Critique. In Scientific Analysis in Archaeology, edited by Julian Henderson. SUCA 19 & UCLA 1A Art 5. Blom, R.G. 1992. Space Technology and the Discovery of Ubar. Point of Beginning, August– September:11–20. Challis, Keith et al. 49 2002. Corona Remotely-Sensed Imagery in Dryland Archaeology: The Islamic City of al-Raqqa, Syria. Journal of Field Archaeology. 29:139-153. Cleave, R.W.L. 1985. Satellite Mapping and the Interpretation of ‗False Color.‘ The Biblical Archaeologist, 48(3):160–1. Cook, S.F. and R.F. Heizer. 1965. Studies on the Chemical Analysis of Archaeological Sites. University of California Publications in Anthropology 2. Berkeley: University of California Press. Cordell, Linda. 1997. Archaeology of the Southwest. San Diego: Academic Press. Custer, J., T. Eveleigh, V. Klemas, and I. Wells. 1986. Application of LANDSAT Data and Synoptic Remote Sensing to Predictive Models for Prehistoric Archaeological Sites: An Example from the Delaware Coastal Plain. American Antiquity, 51:572–88. Davis, J.L. 1993. Interpreting Space: GIS and Archaeology, by K.M.S. Allen, S.W. Green and E.B.W. Zubrow. Reviewed in: American Journal of Archaeology, 97(2): 357–9. Dean, Jeffrey S. 1988. Dendrochronology and Paleoenvironmental Reconstruction on the Color Plateaus. In The Anasazi in a Changing Environment, edited by George J. Gumerman. Cambridge: Cambridge University Press. Deuel, L. 1973. Flights into Yesterday: The Story of Aerial Archaeology, Harmondsworth: Penguin. 50 Donoghue, D.N.M. and I. Shennan. 1987. Remote Sensing of Morton Fen, Lincolnshire: Summary of Recent Work. Fenland Research, 4: 46–53. Donoghue, D.N.M. and I. Shennan. 1988. The Application of Multispectral Remote Sensing Techniques to Wetland Archaeology. In The Exploitation of Wetlands, edited by P. Murphy and C. French, Oxford: British Archaeological Reports. Dorsett, J., D. Gibertson, C. Hunt, and G. Barker. 1984. The UNESCO Libyan Valleys Survey VIII: Image Analysis of Landsat Data for Archaeological and Environmental surveys. Libyan Studies 15:71–80. Du, H., C.-I. Chang, H. Ren, F.M. D'Amico, J. O. J. Jensen. 2004. New Hyperspectral Discrimination Measure for Spectral Characterization. Optical Engineering 43 (8):1777-1786. Fagan, Brian M. 2000. Ancient North America: The Archaeology of a Continent. London: Thames & Hudson. Fiennes, R. 1993. Atlantis of the Sands: The Search for the Lost City of Ur. London: Bloomsbury. Gumerman, G.J. and T.R. Lyons. 1971. Archaeological Methodology and Remote Sensing. Science, 172 (3979):126–32. Gumerman, George J. and Murray Gell-Mann. 51 1994. Cultural Evolution in the Prehistoric Southwest. In Themes in Southwest Prehistory, edited by George J. Gumerman. Sante Fe: School of American Research Press. Hall, Stephen A. 1988. Prehistoric Vegetation and Environment at Chaco Canyon. American Antiquity 53 (3):582-592. Hevly, Richard H. 1988. Prehistoric Vegetation and Paleoclimates on the Colorado Plateaus. In The Anasazi in a Changing Environment, edited by George J. Gumerman. Cambridge: Cambridge University Press. Holliday, Vance T. and William G. Gartner. 2007. Methods of Soil P Analysis in Archaeology. Journal of Archaeological Sciences 34:301-333. Jet Propulsion Laboratory. 2009. ASTER. Electronic document, http://asterweb.jpl.nasa.gov, accessed January 6, 2010. Joyce, C., D. Fuller, and M. Noel. 1992. Archaeology Takes to the Skies. New Scientist, 133(1805):42. Kohler, Timothy A. and Meredith H. Matthews. 1988. Long-Term Anasazi Land Use and Forest Reduction: A Case Study from Southwest Colorado. American Antiquity 53(3):537-564. Larson, Daniel O. and Joel Michaelson. 52 1990. Impacts of Climatic Variability and Population Growth on Virgin Branch Anasazi Cultural Developments. American Antiquity 55(2):227-249. Lunden, B. 1985. Aerial Thermography: A Remote Sensing Technique Applied to Detection of Buried Archaeological Remains at a Site in Dalecarlia, Sweden. Geografiska Annaler. Series A, Physical Geography, 67(1–2):161–166. McHugh,W., C. Breed, G. Schaber, J. McCauley and B. Szabo. 1988. Acheulian Sites Along the ‗Radar Rivers,‘ Southern Egyptian Sahara. Journal of Field Archaeology, 15:361–79. McHugh, W., J. McCauley, C. Haynes, C. Breed and G. Schaber. 1988. Paleorivers and Geoarchaeology in the Southern Egyptian Sahara. Geoarchaeology, 3(1):1–40. McHugh, W., G. Schaber, C. Breed, and J. McCauley. 1989. Neolithic Adaptation and the Holocene Functioning of Tertiary Palaeodrainages in Southern Egypt and Northern Sudan. Antiquity 63: 320–36. Matney, Timothy and Guillermo Algaze. 1995. Urban Development at Mid-Late Early Bronze Age Titris Hoyuk in Southeastern Anatolia. Bulletin of the American Schools of Oriental Research. 299:33-52. Menze, Bjoern H. and Jason A. Ur. 2007. Classification of Multispectral ASTER Imagery in Archaeological Settlement Survey in the Near East. Paper presented the Tenth International Symposium on Physical Measurements and Signatures in Remote Sensing. Palumbo, Gaetano and Dominic Powlesland. 53 1996. Remote Sensing and Geographic Database Management Systems Applications for the Protection and Conservation of Cultural Heritage. In Remote Sensing for Geography, Geology, Land Planning, and Cultural Heritage. Bellingham, Washington: SPIE – The International Society for Optical Engineering. Parcek, Sarah H. 2009. Satellite Remote Sensing for Archaeology. New York: Routledge. Parrington, Michael. 1983. Remote Sensing. Annual Review of Anthropology. 12:105-124. Parry, J. 1992. The Investigative Role of Landsat TM in the Examination of Pre- and ProtoHistoric Water Management Sites in Northeast Thailand. Geocarto International 4:5–24. Provan, D.M.J. 1971. Soil Phosphate Analysis as a Tool in Archaeology. Norwegian Archaeology Review 4:37-50. Provan, D.M.J. 1973. The Soils of an Iron Age Farm Site, Bjellandsoynae, S.W. Norway. Norwegian Archaeology Review 6:30-41 Richards, J.A. 1999. Remote Sensing Digital Image Analysis. Berlin: Springer-Verlag. Russell, E.J. 1957. The World of the Soil. London: Collins. Sever, T. and D. Wagner. 54 1991. Analysis of Prehistoric Roadways in Chaco Canyon Using Remotely Sense Digital Data. In Ancient Road Networks and Settlement in the New World, edited by C. Trombold. Cambridge: Cambridge University Press. Schollar, I., A. Tabbagh, A. Hesse, and I. Herzog. 1990. Archaeological Prospecting and Remote Sensing. Cambridge: Cambridge University Press. Sheets, P. and T. Sever. 1991. Prehistoric Footpaths in Costa Rica: Transportation and Communication in a Tropical Rainforest. In Ancient Road Networks and Settlement in the New World, edited by C. Trombold. Cambridge: Cambridge University Press. Shepherd, Paul R. et al. 1999. The Climate of the Southwest. CLIMAS Report Series, CL1-99. Tou, J. T. and R. C. Gonzalez. 1974. Pattern Recognition Principles. Reading, Massachusetts: Addison-Wesley Publishing Company. Wells, E. Christian et al. 2000. Chemical Analysis of Ancient Anthrosols in Residential Areas at Piedras Negras, Guatemala. Journal of Archaeological Science 27:449-462. Wendorf, F., A. Close and R. Schild. 1987. A Survey of the Egyptian Radar Channels: An Example of Applied Field Archaeology. Journal of Field Archaeology 14:43–63. 55