Short URL

Remote Sensing in Archaeology

 

Other Multi-Sensor Data Fusion areas:

Loading, please wait...

The Practice of Remote Sensing in Archaeology

Geophysical Remote Sensing

Ground-based remote sensing relies on near-surface geophysical remote sensing methods, but with major modifications to methodologies owing to the unique character of archaeological deposits. Modifications to these methodologies include changes in sampling strategies, introduction of raster data structures, and specialized image processing methods. Intensive sub-meter spatial sampling densities are required on the ground in order to enable the possibility of resolving structural features of small size like buried post holes, hearths, storage pits, or walls.


Because large areas must be examined in order to realize the pattern and organization to a settlement's layout, and because high spatial sampling densities are required for the recognition of structural features, a raster data structure is imposed on data collection for uniform spatial coverage. In the field instruments are moved along lines called traverses with measurements, termed samples, acquired at sub-meter intervals depending on an instrument's speed (e.g., 5 cm for GPR, 12.5 cm for magnetometry, 50 cm for resistivity/conductivity). Traverse separation is typically 0.5-1.0 m depending on project goals and the trade-off needs between total area of coverage and the detail and number of structural features detected. Traverses and samples are organized into grids (typically 20 x 20 m in area) to break up large project areas into manageable units. Grids are accurately positioned on the ground by using electronic distance measuring instruments (EDM) and tied to local datums for later co-registration within GIS.

 

Specialized image processing methods are required for data handling. Low-frequency geological trends are typically removed, as is very high frequency spatial variability (e.g. low pass filtering of gradiometer data). The high frequency component that includes human-generated structural features of small size (e.g., 0.25-20 m), is typically enhanced. Periodic features like markings caused by plowing in an agricultural field are typically removed. Attempts are made to remove speckle "noise" that may result from rodent activity or surface plant growth patterns. Contrast enhancement is important to emphasize subtly expressed anomalies. Much emphasis is placed on pattern recognition methodologies following a basic tenet from air photo interpretation: that human constructions tend to produce regular geometries (circles, ovals, rectangles, lines that reflect various architectural features), while most patterns in nature tend to be less regular (with some important exceptions). GPR data are typically subjected to horizontal time-slicing, the generation of a horizontal plan by extracting a slice of data at a particular time (depth) below surface from each profile (traverse). The slices are stacked together to generate a plan image for that depth. Plan views facilitate data interpretation by enabling recognition of structural features by the geometric patterns that are typically associated with them (lines, circles, rectangles). Slices from multiple times (depths) can be acquired allowing visualization of multiple levels of archaeological deposits. Time slicing requires much data processing effort so GPR surveys are typically conducted over small target areas of special interest (e.g., a structural feature suspected to be a "house" from other remote sensing data sets).

 

Three principal geophysical dimensions are examined in archaeology; magnetic, electrical resistivity/electromagnetic conductivity, and ground penetrating radar. Magnetic survey methods are of principal importance in ground-based remote sensing. Magnetic survey methods are passive techniques that are well developed in archaeology, and represent one domain where broad experience has been gained in numerous contexts. These methods are sensitive to subtle soil changes caused by anthropogenic enrichment resulting from human occupation (Clark 1999), any type of soil firing, particularly past the Curie point (about 600o C), and ferrous metal artifacts. Consequently, these methods have proven to be ideal for near-surface archaeological remote sensing owing to the presence of iron artifacts in historical sites, ubiquitous hearths, frequently burned structures, and prior excavations like storage pits, ditches, or house floor depressions that generally fill with magnetically enriched top-soils. A benefit of magnetic survey is its relative speed and high sampling rate (e.g., up to 8 measurements/sec).

 

Human occupations have large impacts on soil magnetism. Anthropogenic enrichment to soil magnetism occurs from the introduction of organic and fired materials. Ditches, pits, cellars and human-caused depressions that are later filled with magnetically enhanced topsoil yield large magnetic measurements. Some ditches or incised trails where magnetically enhanced topsoil is removed are expressed as magnetic voids/lows in data sets. Places of intense firing, hearths, kilns, and burned buildings, markedly increase magnetism and are readily sensed. Ferrous metal artifacts from historic period sites are easily detected.

 

Magnetometry is a relatively fast passive method, enabling high density sampling and large areal coverage. Effective to a depth of 1-1.5 m for many kinds of archaeological features, measurements are an accumulation of the total magnetic response in nanotesla (nT). Current instruments allow data acquisition at 5-10 measurements/second. Sampling at 0.125 x 0.5 m (16 per m2) is common for very high spatial resolutions or with 0.25 x 1.0 m (4 per  m2) common for moderate spatial resolution in a raster data structure. Approximately 0.4 ha can be surveyed per person/instrument per day at very high spatial resolutions, with about 0.8 ha/day at moderate spatial resolutions, assuming open ground without vegetation or any other impediments to easy movement.

 

Magnetic susceptibility measurements, an active method, can now be acquired rapidly in large volumes using new EM instrumentation. Magnetic susceptibility refers to the ability of a soil to be magnetized allowing the mapping of the potential of deposits to be magnetized due to anthropogenic activity. While passive magnetometry records the total magnetic field or its gradient, magnetic susceptibility instruments employ active methods to induce a temporary magnetic field allowing quantification of magnetically enriched deposits created by anthropogenic activity. Such information can yield additional insights about the nature of settlement boundaries, midden deposits, and overall site structure (Clark 1999). Recent advances in instrumentation allow rapid acquisition of these data at high spatial resolutions. Prospecting depth is probably less than 0.5 m. Sampling density is typically on the order of 0.5 x 0.5 m to 1 x 1 m in a raster data structure. About 0.4-0.8 ha can be surveyed per day depending on sampling density.

Magnetometry

Electrical resistivity/Electromagnetic conductivity (EM) instruments involve active methods that measure soil conductivity (where resistivity is the inverse of conductivity). Electrical resistivity is an active method applicable to nearly all soil environments. It is responsive to subtle contrasts, rapid soil changes, and can be targeted to specific depths. It is very useful in delineating soil changes caused by prehistoric house depressions, ditches, midden deposits, and trails, for example. Building stone and brick, common to house foundations of the historic Euroamerican period, exhibit large resistivities and are readily sensed. While it has only recently gained a large acceptance in North American archaeology (Kvamme 2001), resistivity survey has a long tradition in northern Europe (Clark 1999). Recent instrumentation developed by Geoscan Research (a project participant) that is optimized for archaeological applications, allows extremely rapid data collection over broad areas.

Soil conductivity is an active method that is the theoretical inverse of resistivity. Yet very different non-contact instruments operating with radio frequencies in place of currents injected through probes are utilized to assess soil volumes. Results can therefore be very different from resistivity (Kvamme 2001). Different deposits exhibit variations in soil conductivity due to changes in soil moisture, parent material, compaction/porosity, and other factors, measured in milliSiemens/m or ohm/m. A large component of the mapping of soil conductivity reflects variations in soil moisture. Rocks, stone or brick foundations, and pavements tend to exhibit very high resistivity, while the sediments that have filled in ditches, pits, and other depressions tend to retain greater moisture that lowers resistivity. Moreover, EM instruments are sensitive to highly conductive metallic artifacts, a feature that can be useful at historical sites. The most common EM instrument used in archaeology averages soil conductivity through a volume of 1.5 m or 0.75 m below surface, depending on the instrument's operation mode. The archaeological community has less experience with soil conductivity instruments, but recent studies and new instrumentation hold great promise.

 

Electrical resistivity measurements can be acquired at more specific target depths to, potentially, about 2 m. Because no probes are used EM data acquisition is marginally faster, although recent electrical resistivity designs with multi-probe arrays, multiplexers, and rigid frames enable very fast surveys when certain probe separations (depths) are employed. Sampling with both methods typically varies between 0.5 x 0.5 m to 1 x 1 m in a raster structure, depending on speed/resolution trade-offs and the "crispness" of the signal (sometimes denser sampling does not improve the result when soil conductivity changes are relatively slow). Experience suggests that electrical resistivity surveys may reveal somewhat better detail (e.g., edge definition) owing to the use of probe placements and the ability to more closely target specific depths.

Electrical Resistivity



 

Ground penetrating radar (GPR) is an active method that responds to subsurface dielectric changes between deposits. GPR is a relatively new technology designed and developed primarily for geotechnical applications (high contrast targets), but has nevertheless found application in many archaeological environments, often with excellent results. Most GPR systems used in archaeology send continuous pulses of radar energy in the 300-500 MHz range vertically into the ground along the full length of a survey transect. These pulses reflect off buried features such as stratigraphic interfaces, walls, house floors, or pits. The return times of echoes from these pulses give information on depth, and their magnitudes indicate something of the nature of the subsurface reflectors. The outcome mimics a section or profile along the length of the survey line. Thus, GPR data in their native form are ideally suited for gaining information in the vertical plane, including stratigraphic relationships (Conyers and Goodman 1998). Recent advances in GPR data processing allow generation of horizontal "time slices" between the profiles, providing a means to generate plans that greatly facilitate interpretation and correlation between anomalies in adjacent profiles (Goodman, Nishimura, and Rogers 1995). GPR yields data from multiple "depths" simultaneously, where depth is represented by travel times from the antenna in nanoseconds. Horizontal "time-slicing" between profiles generates multiple raster data sets in plan view, from various times or depths below surface. Changes in dielectric properties between deposits cause reflections of GPR pulses to be returned to the surface; larger contrasts in the relative dielectric permitivity cause larger amplitude reflections. GPR is therefore good at mapping changes in subsurface deposits and materials detecting virtually any kind of structural feature. High antenna frequencies (300-900 MHz) typically allow prospecting depths to 1-4 m depending on soil conditions with good spatial resolution of structural features sub-meter in size, especially with higher frequencies. Low frequency antennas (< 300 MHz) allow deeper prospecting but spatial resolution is poor. Signal attenuation or dispersal can occur very rapidly below the surface when soils are highly conductive (e.g., moist clays).

 

The native form of GPR data is a raster, but represents a vertical profile or section, with the horizontal axis distance along a traverse and the vertical axis two-way travel times of the pulses beneath the surface. Spatial sampling along a traverse or profile is typically high (1-5 cm) with profile separation is typically set at 0.5 m for high resolution surveys to 1.0 m for moderate resolution. Because sampling density is so much higher GPR provides the greatest probability of detecting structural features of small size. Areal coverage/day is typically low, perhaps 0.2-0.3 ha/day although greater amounts could be performed if data processing was not so cumbersome. Data processing is relatively slow compared to other methods and data sets are very large, with specialized algorithms required. For this reason GPR is typically applied to certain target areas of special interest, although it will be employed for wide-area survey in the proposed project.


Ground Penetrating Radar (GPR)

 


Aerial Remote Sensing

 

Aerial remote sensing methods record vegetation, soil, moisture, and temperature changes. Historical and modern panchromatic aerial photography will be acquired and scanned into a raster format to record surface indications of buried archaeological features and to document recent changes to sites. Aerial photography is typically acquired in black-and-white or color panchromatic, or in color near infrared (NIR). Aerial photographs, rectified and registered to a common coordinate base, typically yield sub-meter spatial resolution, depending on the photo format, aircraft altitude and speed.

 

Characteristics of archaeological sites are revealed in the air by several phenomena. Crop or vegetation marking results from differential growth patterns caused by underlying archaeological deposits where structural features composed of rock typically impede, stunt, or delay growth, while beneficial soils like moist sediments filling ditches, pits, or depressions tend to enhance or advance growth patterns. Snow or frost marking results from thermal properties within the deposits that cause differential snow or frost melt, which reveal microtopographic relief orientations stemming from the archaeological deposits. Microtopographic relief variations can also be made visible by shadowing in low sunlight angle conditions. Moisture, color, and texture variations of soils in freshly plowed fields cause soil marking that can reveal near-surface archaeological content.


Panchromatic Photography


 

The Advanced Thermal and Land Applications Sensor (ATLAS) is a 15-channel multispectral scanner that incorporates the bandwidths of the Landsat Thematic Mapper with additional bands in the middle reflective infrared and thermal infrared range (TIR). Spatial resolution of the sensor is a function of aircraft altitude. For the purpose of this project, existing (but as yet unstudied) data at a 5 meter resolution from Fort Bliss will be used. Of particular importance to the detection of archeological features is the multispectral TIR capability of the ATLAS instrument, which permits the accurate measurement of thermal responses for different landscape characteristics. These thermal bands have successfully detected ancient Anasazi roadways and agricultural fields in Chaco Canyon, NM, and prehistoric footpaths in the Arenal region of northern Costa Rica (Sever 1990). This method is illustrated in with data from Chaco Canyon. All manifestations of archaeological sites seen in aerial photography should be detectable to the limits of the obtained spatial resolution, plus many more owing to the greater spectral breadth and control of spectral resolution.


Aerial Imagery

 

One of the results of NASA's scientific and commercial remote sensing programs has been the development of lightweight, low-cost portable thermal sensors. The Inframetrics Model 740, originally employed for ice-detection on the shuttle, is a hand-held instrument that has a sensitivity to 0.1o C and can be used at ground level or installed in a helicopter or fixed-wing aircraft. The Agema 880 is a dual band IR image scanner. The Palm IR 250/255 is a single band hand-held scanner. These instruments have great potential for archeological research and were used to detect the footprint of the foundation of the 1910 Wright-Brother Hangar (Sever 1995), graves and other cultural features at a Confederate prison camp in Salisbury, NC (Real Time Thermal Imaging), and a 200 year old water ditch, buried rock pathway, and 14 Indian burial trenches from the small pox epidemic of 1838 (Larimer 1988). Both portable rough-terrain aerial boom lifts and rotary wing platform acquisitions will be evaluated. At installations where boom lifts are available the site will be imaged repeatedly from before dawn to late morning. The acquisition of this type of thermal data set will allow us to investigate, in detail, the results of differential heating, which may be a strong indicator of different features within a site.

Georeferencing is a limitation of hand held sensors for data fusion. Co-registration (particularly of off-nadir data sets) presents particular challenges in photogrammetry. Small image size may require multiple images mosaics to cover area of interest. Approaches to overcome limitations include the use of modern soft bench photogrammetric software, which allows the creation of ortho-corrected raster products from off-nadir data and non-metric camera sources based on extensive, thermally-visible ground control points (GCPs). Team members (CAST) have been active in development of these solutions for archaeological applications (Gisiger et al 1998).


Thermal Imagery


 
Satellite Remote Sensing

 

The IKONOS high resolution satellite was launched in September 1999, and has two imaging sensors: the panchromatic and multispectral. IKONOS provides one meter resolution panchromatic imagery and four multispectral bands at 4 m resolution. This satellite has a polar, circular, sun-synchronous 681-km orbit and both sensors have a swath width of 11 km. IKONOS data has been analyzed over the Peten region of northern Guatemala. Features not apparent in the Landsat TM imagery are easily visible in the IKONOS data as indicated below. QuickBird was launched on October 18th and appears to have achieved stable orbit and system health. Data are not yet available. QuickBird's specifications provide for 0.61 m panchromatic and 2.44 m four-band multispectral data. Pricing has not yet been published. Neither of these two high resolution data sources has yet been extensively field tested for archaeological use. Both offer 11 bit (2,048 gray levels) resolution. New image classification methodologies now permit the spatial detail of the 1 or 0.61meter band to be fused with the spectral data of the 4 or 2.44 meter band prior to classification providing a 1 or 0.61 meter multispectral capability. Sub-pixel classification techniques also allow the identification of phenomena on the surface with a sub-meter spatial extent. All manifestations of archaeological sites seen in aerial photography (see above figure) should be detectable to the limits of the obtained spatial resolution, plus many more owing to the greater spectral breadth and control of spectral resolution. Ortho-photogrammetric methods are employed for image rectification and registration to a common coordinate base.


Space Imagery