SUB PIXEL DETECTION OF OBSIDIAN AND POTTERY BY NASA SATELLITE AND AIRCRAFT DATA

We determine the detection limits of sub-pixel archaeological artifacts using airborne and spaceborne image data. Research results are presented from the Glass Mountain Site in northern CA and the Boquillas site in southern AZ. Multiple visits were made at different seasons over 3 years. The average density of site midden, obsidian artifacts, and pottery (per m²) has been calculated for 25 8 x 8 m squares and 10 4 x 4 m squares. Spectral reflectance data were measured in the field using a portable field spectrometer (ASD 350-2500P model) in visible, shortwave infrared and thermal infrared of targets and backgrounds. A spectral reference library has been constructed of ~200 target and background samples. Image data include NASA’s MODIS/ASTER airborne simulator (MASTER) imaging system, the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER).

We analyzed images of the study areas using spectral mixture analysis which characterizes mixed pixel spectra as linear combinations of reference spectra and shade. This approach incorporates the dominant factors that affect detectability, including: 1) the spectral contrast between the target and background materials, 2) the proportion of the target on the surface (relative to background), 3) the imaging system being used (bands, instrument noise, pixel size), and 4) the conditions under which the surface is being imaged (illumination, atmosphere). The surface area on the ground, represented by an image pixel, must have at least a certain fraction of exposed target to be detected. That fraction is the “detection threshold.”

For our research, we 1) determined the detection limits of obsidian and ceramic artifacts at the sub-pixel scale; 2) examined the influence of background, seasonal vegetation change and other on-site changes on the detectability of these artifact types; 3) established the instrumentation, spatial scale, and spectral bands needed to improve the detectability, and 4) tested predictions of new locations for obsidian artifacts at specific (spatial) densities in other image scenes and ground truthed these predictions.