Radar Archaeology: Space Age Tools Aid in Uncovering the Past
By Derrold W. Holcomb

Use of remote sensing technology for problem solving is dictated by the economics of the application. In the first years of image processing, the neccessity for expensive mainframe computers limited use of this technology to such high-dollar industries as oil and mineral exploration, large-scale resource management or military applications. In the late 1970s ERDAS Inc. pioneered the idea of image processing on small personal computers. The resultant dramatic decrease in hardware costs, along with a phenomenal increase in computing power since that time, have revolutionized the field, making the technology ever more accessible. Similarly, with the recent launch of many new satellites the data itself has become more affordable and spatial resolution is increasing. There are currently more than a dozen new sensors on the drawing boards, promising resolutions as high as 1 to 2 meters.
      One field in which satellite imagery is now becoming a cost-effective solution is archaeological reconaissance and survey. In addition to collecting artifacts and cataloging physical traces of past human activity, contemporary archaeologists are concerned with a geographic and conceptual overview on a much larger scale, studying the processes of change operant in mans history and prehistory. Important in this study is the relationship between human society and the environment. The expanded perspective afforded by space data allows study of regional sequences and large area phenomena, including the impact of mans presence on the landscape.
      Beyond direct detection of archaeological sites, remotely sensed data offers a means of indirect detection, defining areas where human activity is likely to have occurred. Regional surveys on a scale not possible from the ground can help ameliorate the time consuming and expensive work of conventional archaeological reconnaissance and survey. This is particularly valuable in inaccessible areas. Remote sensing techniques are well-suited for investigation of such factors as availability of water, access to arable land, sources of raw materials or likely routes of transportation and communication.
      A particularly significant attribute of the radar sensor for archaeological applications is its penetration capabilities. A major problem in any remote sensing study, particularly in tropical or sub-tropical climates, is obstruction of the image by cloud cover. The development of radar sensors has obviated this difficulty, as clouds are essentially transparent at radar wavelengths. Sensor research indicated that the longer wavelength sensors may also be able to image through vegetation. One of the more surprising findings, however, was the ability of the radar to detect subsurface features in arid regions.
      In 1981 the Space Shuttle Columbia carried an experimental synthetic aperture radar sensor, Shuttle Imaging Radar-A (SIR-A), which collected images from numerous areas worldwide. Figure 5 shows intersecting images from the SIR-A sensor and the Large Format Camera (LFC), a high resolution panchromatic camera carried on the same mission. It is immediately apparent that the radar image shows more detail than the visible light image from the LFC in this desert area (Figure 1). The radar signal is able to penetrate the overlying sandsheet to reveal fine morphologic structure beneath.
      In data from the SIR-A mission, the first radar rivers were detected in the southeastern Sahara Desert. A dendritic system was noted on radar images that did not appear on visible wavelength Landsat images of the same area. Fieldwork indicated that the radar sensor had mapped an ancient drainage system buried under layers of dry sand up to 2 meters in depth. The courses of ancient waterways through desert areas are more densely consolidated than surrounding or overlying sand, thus yielding a discernibly different radar return. Based on the probability of association between water sources and human activity, the capability to discern such subsurface features holds great potential for archaeological applications.
      Under the proper conditions the radar burst from an imaging radar sensor will penetrate loosely consolidated sand to produce an image reflected from a denser medium beneath the sand cover. In order to be detected, an object need not physically fill the entire pixel area, but need only dominate the return signal. Strength of return signal is affected by many factors, such as reflectivity and geometry of an objects surface. For example, a highly reflective pipeline, oil derrick or railroad track whose actual dimensions are far smaller than a single pixel is easily imaged. The geometry of a linear feature such as a wall running perpendicular to the line of sight from the sensor would produce a strong return signal. The same wall running parallel to the line of sight could be invisible to the sensor. In nature right angles and systems of parallel lines are rare, therefore generally implying human activity. Rectangular buildings, defense walls, roads, irrigation channels or orchards are all potentially imageable features.
      If it is possible to image sand-covered rivers, might it not also be possible to detect remnants of sand-covered cities? This possibility, to image mud walls covered by drift sand, presses the limits of the radar sensor, both spatially and radiometrically, thereby making it a perfect test case for software development. In developing its Radar Module, ERDAS included this archaeological limits-of-detection problem into the testing regimen. Enhancement and despeckle algorithms have been developed to retain detail in subsurface images at the single pixel level. These are the foundation for a growing repertoire of enhancement regimens designed specifically for archaeological applications.
      These enhancement regimens were cautiously deduced by observing the effect of each operation on pixels known to be of interest. The radar rivers located in Egypt were hundreds of meters in width and discernible on unenhanced imagery. The braided stream indicated by the arrow in Figure 1, however, is much smaller, at times no wider than 1 pixel (25 meters). Resolution at this scale is confounded by the speckle inherent in imaging radar systems. An initial pixel averaging operation to smooth the digital data is a common preprocessing step for radar data. It was found, however, that such preprocessing resulted in a total loss of the narrow streambeds in Figure 1. Sequences of operations have been developed, yielding enhancement regimens capable of elucidating features of interest without loss of fine detail or accentuation of speckle artifact. Pursuit of this archaeological line of inquiry culminated in an experiment on NASAs SIR-C mission, a radar search for lost cities in Central Asia.
      The Silk Road of antiquity was the major interface between the predominant cultural centers of 2,000 years ago, including Chinese, Greco-Roman, Arab and Indian peoples. Remains of cities and settlements at oases along the ancient trade route have been extraordinarily well preserved in the hyperarid desert environment. As desertification has advanced, sites which once were watered by oases along the deserts edge now lie far out into hyperarid regions.
      In 1899, Aurel Stein crossed the Himalayas from British India on his first trip into Chinese Turkestan. The foremost of a number of western archaeologists and explorers coming through Russia or the Himalayas into Chinas northwest desert, he brought back historic treasures from the famed Silk Road. After careful documentation, Stein frequently reburied his excavations, leaving the sites preserved beneath the desert sands. Many of these finds have since been lost, swallowed again beneath the shifting sands of the expanding desert at the foot of the Tibet Plateau. Until very recently, all that remained of many significant archaeological sites were museum collections of artifacts, photos and Steins hand-drawn maps. In 1994, a new era of exploration was inaugurated in this area when the space shuttle flew over the Himalayas. One of the experiments for the SIR-C mission was an attempt to relocate some of the Silk Road archaeological sites Stein had found a century earlier.
      For this investigation, the entire archive of Steins work was carefully researched, including field notes, photographs and, most importantly, his original hand-drafted survey maps. While his archaeological maps show high local precision, the sites are not accurately located on a larger scale, due to the technical limitations of overland navigation at the time. His survey was generated with a theodolite, wheel roller and painstaking notes carried overland for hundreds of miles across the Himalayan Mountains and the relatively featureless Taklamakan Desert. His estimates of geographic location could not provide the accuracy required for any hope of success in targeting the space shuttle sensor. The Taklamakan is the worlds second largest desert, encompassing 338,000 square kilometers. Sensor imaging areas are generally less than 150 km square, so highly accurate targeting information is required to obtain usable data. Therefore, Steins turn-of-the-century maps had to be updated.
      Study areas in the region of Khotan Oasis were selected based on historical evidence of the existence of significant archaeological sites and environmental factors favorable to remote sensor imaging. In 1992, a two-month expedition was mounted to the study area, in what is now Chinas Xinjiang Province. With a GPS receiver, navigation satellite fixes were taken for sites such as river junctions, town centers and historic mosques, locations likely to have remained unchanged over the years. Accurate latitude and longitude information was then used to rectify Steins original maps and generate sensor targeting coordinates calculated to optimize the chances of imaging areas containing lost archaeological sites. The resultant targeting information was submitted to NASA for the recent SIR-C mission.
      Results from this mission proved these efforts worthwhile. Computer enhancement of the digital radar imagery has revealed linear features interpreted as a man-made dike and canal system which may once have directed irrigation water toward the ancient Silk Road settlement of Niya. Figure 4 shows a composite of SPOT panchromatic and SIR-C radar imagery. The two frequencies and two polarizations of the multi-spectral radar image yield a wealth of information conveyed as color differences. This study is an excellent example of the value of combining data from two very different sensors. The feature of interest, detailed in Figure 2, was first detected as an anomaly in the radar imagery. As areas seen in white were interpreted as evaporite deposits, why would a line of white transect a ridge? To call this feature a man-made canal would have been too speculative, based on the radar data alone. The theory gains credibility, however, in combination with the SPOT image, in frequencies similar to those seen by the human eye, and is further supported by the archaeological record. Historic agricultural settlements have been documented in the vicinity of the dry waterways imaged, hundreds of kilometers downstream from where modern rivers disappear at the boundaries of the expanding desert.
      The next phase of this on-going research involves fieldwork on site to verify these interpretations and to gather data for further refinement of the sensor and applicable computer enhancement regimens. It is expected that, by using satellite images and GPS, a field team will be able to navigate accurately through this extreme desert terrain, facilitating detailed archaeological survey and/or excavation.
      In another current example, space archaeology is proving a useful complement to excavation of the ancient oasis of Merv, in Turkmenistan. Like the Niya site, Merv has been designated a UNESCO International Archaeologic Cooperation site and, as such, was also imaged on the SIR-C mission.
     Radar imagery is being used in a search for ancillary sub-surface features associated with the massive city walls of Merv, still standing in the desert. Again, the synergistic use of radar and visible light imagery has proven advantageous. As vegetation patches in a desert environment are of archaeological interest, a vegetation index image (NDVI) was computed from SPOT multi-spectral data and compared with the L-band radar image in Figure 3. Areas of four distinct radar return signals are labeled in the image. The mottled areas labeled D seem to correlate with the NDVI and were tentatively assumed to be due to radar scattering off vegetation, (a known phenomenon). The bright arc labeled B falls directly on the walls of the circular citadel. This and the series of parallel lines north of B are all oriented perpendicular to the radars southwest look direction, a favorable geometry for generating a strong return signal. These were therefore assumed to be NW-SE trending surface features. The bright returns at A and C, however, could not be explained at the image processor, and were targeted for further investigation on the ground in this years fieldwork.
      At what is presumed to be the entrance walls of the eastern gate into Merv, dirt ridges were found perpendicular to the city walls, forming a perfect corner reflector which would explain the very bright signal return at A in the radar image. The bright return at C was generated from the northeast side of a huge mound within the citadel, presumably a palace or important building. Other anomalous areas found within the digital database were investigated on the ground, some of which were found to be remnants of isolated structures in the surrounding desert. In continuation of this ongoing work, a NASA-RADARSAT grant is supporting development of a paleohydrologic map of Central Asia which should contribute to further archaeological survey in the region.
      An important issue facing the fledgling field of space archaeology is how to bring this new technology to archaeologists worldwide. Remotely sensed data and computer processing are expensive. Large scale enterprises such as oil and mineral industries, fisheries and global environmental monitoring concerns have already found remote sensing to be a good investment. Now, as more satellites are launched and data becomes more affordable, archaeologists can begin developing reconnaissance applications to optimize field efforts on the ground. It should become feasible to survey large areas in advance, to better direct the costly and time consuming process of ground exploration. In addition, some archaeological features (the Nazca lines being the most famous example) are difficult or impossible to discern at close range, becoming visible only on the larger scale afforded by viewing the earth from space.

Acknowledgements:
Thanks are due to NASA-JPL for all radar imagery presented here. SPOT Image Corp. kindly donated the two SPOT images. ERDAS Inc., has supported this work for many years.

About the author:
Derrold W. Holcomb is a consultant specializing in applications and algorithms for advanced sensors and is currently employed at ERDAS Inc. in Atlanta, Ga. He may be reached at 404-248-9000.

Back