The research reported herein focuses on developing and testing predictive models based on the satellite remote-sensing (SRS) of prehistoric and historic archaeological phenomena. With advances in the resolution of satellite-borne imagery, such as IKONOS, and the availability of software designed to process such imagery, such as ENVI, archaeological predictive modeling is positioned to progress beyond simplistic “indirect” correlational studies involving gross ecological categories or subjective landform designations.These developments are particularly important because previous applications of SRS have focused principally on comparatively uncommon large-scale and obtrusive
archaeological remains, thereby constraining the range of SRS applications and disincentivizing the exploration of the technology’s predictive modeling capabilities. To take advantage of these technologies, we investigate the potential of high-resolution SRS of small-scale and low-obtrusive archaeological phenomena from the Upper Basin of northern Arizona (Kaibab National Forest, Tusayan Ranger District) to develop a new approach to predictive modeling.
We illustrate how the GPS-determined locations of the Upper Basin’s masonry structures, brush structures, fire-cracked-rock piles, lithic scatters, and sherd-and-lithic scatters can be overlaid on the pixels of a geo-registered, clipped, and masked IKONOS image to yield distinctive spectral signatures. In this case, variation among these five types of archaeological phenomena was sufficiently robust, principally with respect to their reflectance properties, that they registered very differently with the IKONOS satellite sensors. The reliability of a predictive model based on these SRS-based spectral signatures was subsequently tested in newly surveyed terrain. The model correctly predicted the presence and the absence of the five types of archaeological phenomena at extremely high rates (100% and 99.4%, respectively).
One major methodological finding of this study is that the resolution of pixels needs to be adjusted from the default imagery settings in order to avoid producing an over-abundance of false positives. In other words, failure to degrade the SRS image from 1m to 4m will result in heavy environmental mimicry of archaeological phenomena, which essentially over-predicts their regional frequencies (leading to inaccurate predictive estimates). Additional research is needed to understand how “tweaking” of pixel resolution affects the derivation of spectral signatures of different archaeological phenomena and the extent to which they then can be accurately estimated on a regional basis.
As these “direct” predictive models are refined they will become increasingly more useful in historic preservation and heritage management, especially with respect to identifying archaeologically sensitive areas that require special management actions, such as dynamic monitoring, protection, or exclusion.
This research was made possible through Grant MT-2210-05-NC-12 from the National Center for Preservation Technology and Training (NCPTT).