SUB-PIXEL DETECTION OF ARCHAEOLOGICAL MATERIALS USING NASA SATELLITE AND AIRCRAFT DATA

For this NASA-funded project we examine whether sub-pixel artifacts (i.e. site midden and concentrations of obsidian artifacts and pottery sherds ) can be directly detected/identified using airborne and spaceborne image data. The objectives of the research are to: 1) use NASA image data in conjunction with actual field/laboratory measured spectra of archaeological materials to test the detection limits of selected artifact classes (obsidian, ceramics and midden) at the sub-pixel scale by applying previously demonstrated theoretical detection limit modeling, 2) examine the influence that background, seasonal vegetation change and other on-site changes have for the detectability of these objects in image data, 3) establish the instrumentation, spatial scale, and spectral bands needed to improve the detectability of these objects, and 4) to test predictions of new locations for artifacts at specific (spatial) densities in other image scenes and ground truth these predictions. We are investigating two locations in western North America with common concentrations of obsidian (Glass Mountain, CA) and pottery/midden (lower San Pedro River Valley, AZ). NASA data in visible and TIR are the primary image data used, available from aircraft and satellite at varying scales over these archaeological sites. Spectral characteristics of targets and backgrounds will be measured in the field and laboratory, and a mixture model constructed linking these spectra to image data. The success will be evaluated by mapping predicted concentrations nearby and conducting ground truthing to determine accuracy. We make intensive use of NASA data especially that collected from MASTER, ASTER and AVARIS instruments. This research will give archaeologists and prehistorians a new way of using NASA data to examine the archaeological record, and may provide land managers in the US with new tools for identifying and protecting cultural resources, a clear societal benefit.