Subpixel artifact detection using remote sensing

Affiliation(s) PI Project Period Funded by
DEES Buck, Paul 2000-Present U.S. Department of the interior, Bureau of Land Management

Keywords: remote sensing, subpixel artifact detection

Project Description

This project explores whether certain kinds of common archaeological artifacts are detectable on the earth's surface using remote sensing, especially thermal instrumentation. Using a mixture modeling approach developed by the University of Washington's Keck Geological Remote Sensing Laboratory, Buck and UW colleagues Sabol and Gillespie have shown a number of interesting and potentially useful results.

In a prior study (Buck et al. n.d.), the sub-pixel spectral detectability of obsidian and ceramic artifacts against typical soil backgrounds from two study areas in the western U.S.A. was analytically evaluated to determine the usefulness of remote sensing as a tool for artifact detection [in the visible (VIS), near-infrared (NIR), and/or thermal-infrared (TIR) portion of the spectrum]. In the VIS/NIR, surface concentrations of pottery needed to be 85% or greater to be detected against backgrounds of soil, rock, and vegetation. At the same wavelengths, obsidian is spectrally similar to shade and cannot be uniquely detected unless the effects of shade are independently removed. In cases where shade is not a major factor, obsidian at concentrations of 2-3% can be detected. In the TIR, pottery thresholds ranged from 12-48%, depending on the composition of the background. Obsidian detection thresholds ranged from 4% to 25% cover. These results indicate that surface mapping using remote sensing in the TIR has the potential to be an effective tool for prioritizing large areas for future ground surveys.


spectral image
spectral image