Finding Archaeological Sites Using Airborne Radar
a case study at san clemente island, california using synthetic aperture radar (sar)
What is Sar?
Synthetic aperture radar (SAR) is a method of transmitting and receiving radar at different frequencies, or bands, from a moving platform such as an airplane or satellite. Collecting radar data in this way improves resolution, which allows SAR platforms to cover wider areas at faster rates and at higher levels of precision compared to other radar platforms such as ground penetrating radar (GPR).
As with conventional radar, SAR is used to study the physical and structural attributes of the area one is interested in. Radar at any given frequency and polarization penetrates different materials in different ways. If a SAR sensor is capable of collecting data from multiple bands, each band in particular is suited for measuring different aspects of a landscape such as topography, electrical permittivity (the measurable responses of a material to electric fields), roughness, and geometry (structure). SAR is able to penetrate features such as clouds, vegetation, and soil to obtain readings where optical imaging cannot "see".
Archaeological site detection at san clemente island
Over the course of several decades, archaeologists have recorded all archaeological sites accessible to pedestrian survey on San Clemente Island. Due to this fact, San Clemente served as a prime testing ground for the use SAR in detecting archaeological sites.
CSRM, together with NASA's Jet Propulsion Laboratory (JPL), utilized SAR technology to detect archaeological sites at San Clemente Island. Two platforms were used: GeoSAR and the AirSAR platform mounted on a DC-8 aircraft, which collected three SAR bands (P-, L-, and C-band) polarized in several different ways.
The different bands and polarizations of these SAR platforms were able to detect different aspects of archaeological sites on San Clemente. Here, most archaeological sites contained buried middens, or trash heaps, that contained dark, organic, and electrically conductive soils which differed significantly from the surrounding native soils. In addition, these features commonly contained scatters of fist-sized rocks and a high amount of vegetation. All of these characteristics yielded unique signatures in the SAR imagery. The P-band, which is normally unaffected by most material, was reflected by the moist, electrically conductive midden soils. This created a noticeably weak signature for the middens in the imagery compared to the surrounding environment. The shorter L- and C-bands were also scattered in different ways by rock scatters and vegetation. For example, the grasses present atop the middens generated strong returns from short bands polarized vertically at both transmission and reception, but showed very weak returns when transmitted vertically and received horizontally (and vice-versa).
These backscatter readings from the known archaeological sites were compared to areas that were known to contain no archaeological sites. The Student's t-test was performed to show that the readings of sites vs. non-sites for several polarized bands was statistically different to a 95% degree of certainty. These results were then used to create a model showing where 0, 1, 2, or 3 bands detected archaeological sites across the island (seen at the top of the page).
Gain statistics (1- % area / % sites) (Kvamme 1988) were also calculated to see how productive the model was in detecting archaeological sites. A maximum gain statistic would be 1.00 while any negative number would indicate a poor correlation. As seen in the table above, more than 75% of San Clemente Island corresponds to a large negative gain statistic for the model, indicating that extremely few recorded sites were present in parts of the model where no sites were indicated. At the same time, as more bands indicated the presence of archaeological sites, the gain statistic dramatically rose close to 1.00.
The results of this study show that site detection using SAR is incredibly productive for mapping the location of archaeological sites. It should be noted that more "traditional" methods for predicting archaeological sites, such as those done by associating environmental factors with site locations, have rarely, if ever, exceeded a gain statistic of 0.70 (Ebert 2000: 133). Site detection models using SAR could be of tremendous use both for academic research and for those managing protected sites. Site managers and custodians could use these models to indicate where development would adversely effect archaeological resources. This in turn would allow decision makers to determine where development should be avoided, or if development is unavoidable, how to most efficiently test for and recover archaeological materials.
References and Further Reading:
Comer, D.C., R.G. Blom, and W. Megarry. (2013). The Influence of Viewshed on Prehistoric Archaeological Site Patterning at San Clemente Island as Suggested by Analysis of Synthetic Aperture Radar Images. In D.C. Comer & M.J. Harrower (Eds.), Mapping Archaeological Landscapes from Space (pp. 159-171). Springer: New York.
Ebert, J. (2000). The State of the Art in "Inductive" Predictive Modeling: Seven Big Mistakes (and Lots of Smaller Ones). In K.L. Wescott & R.J. Brandon (Eds.), Practical Applications of GIS for Archaeologists: A Predictive Modeling Toolkit (pp. 129-134). Taylor and Francis: London.
Kvamme, K. (1988). Development and Testing of Quantitative Models. In W. James Judge & L. Sebastian (Eds.), Quantifying the Present and Predicting the Past: Theory, Method, and Application of Archaeological Predictive Modeling (pp. 325-428). US Department of the Interior, Bureau of Land Management: Denver.