The Department of Defense (DoD) manages 30 million acres of public land containing some of the nation's most significant historic and prehistoric cultural resources. Protecting these heritage resources is a fundamental part of DoD's mission. Heritage management issues central to that mission have focused on the economics of identifying and maintaining historic facilities, the impact of archaeological sites on construction and training programs, and the disposition and curation of artifacts. Management of these resources, in compliance with existing laws and regulations, necessitates the development of innovative and cost-effective methods for archaeological site identification, evaluation, and protection; however, archaeological methods for the identification and evaluation of most historic resources remain essentially unchanged since the early twentieth century. Surface survey and excavation--the traditional field methods for discovery of artifacts, architectural elements, and other features--predominate in spite of the fact that these techniques are extremely time consuming, expensive, and unreliable. What is needed are methodologies for "seeing" into the ground and detecting subsurface archaeological deposits over large areas. Integration of multiscale multiplatform remotely sensed data offers an opportunity to recover a great deal of information about archaeological site content while reducing costs associated with field work and long-term curation of excavated collections.

The objective of this project was to acquire and analyze a suite of ground, aerial, and space-based sensor data to determine which remote sensor combinations, data fusion techniques, and analytical approaches most accurately predict the presence and specific nature of subsurface archaeological features in various environmental conditions and archaeological settings. Follow-up archaeological field testing provided a quantitative assessment of each approach's ability to successfully detect buried features. The archaeological field validation studies, patterned regularities in the data, modality correlation, and other factors were used to evaluate sensor performance and the effectiveness of data fusion methods and analytical approaches.

The project began with the acquisition/compilation of geophysical data--magnetometry, magnetic susceptibility, electrical resistivity/conductivity, ground penetrating radar, terrestrial thermal infrared, aerial photography, and satellite-based, high-resolution multispectral (visible, near-infrared)--for a wide range of archaeological site types at Fort Benning, Fort Bliss, Fort Riley, and the Department of Energy's Savannah River Site. After data preprocessing, a range of multiple, alternative methods of fusion and analysis were applied for the detection of archaeologically relevant evidence (foundations, hearths, etc.). Categories of fusion included multiband visualization, pixel-based multivariate statistical methods, image segmentation and object classification, and model-based fusion. The fusion methods applied a range of pattern/structure recognition approaches to the problem of content extraction from multisource data. In the first technique the recognition process was totally visual, in the second statistical methods were used to identify patterning, in the third method initial statistical methods were joined with user supplied rule sets, in which the rule set reflects a priori knowledge about the spatial relationships of the statistically derived object primitives. In the final model-based data fusion approach, a priori information on instrument values and physical properties were built into the process models that then were calibrated through iterative application to data. The anomalies detected in the data fusion/analysis phase were field tested using established archaeological excavation techniques to quantify prediction accuracy by recording the ratios of detected features to false positives and to false negatives. The resulting assessment allowed the performance of the various sensing methods to be evaluated individually and together. The utility and benefits of the various data fusion/integration methods also were quantitatively measured and assessed.

The inclusion of fused remotely sensed data analysis in the early stages of archaeological site assessment represents an alternative approach with the potential to overcome some of the typical limitations of both traditional (non-remote sensing) site evaluation and single sensor surveys. Multisensor surveys permit the detection of features across large portions of sites, improving the reliability of site assessments. The benefits of improved site detection are threefold. First, thorough evaluations based on hand excavation are expensive. Remote detection of cultural resources reduces the amount of excavation needed to effectively evaluate a site, significantly reducing the associated labor and logistics costs. Second, reliable location of archaeological features and attendant reduction of excavation reduces the volume of artifact collections, resulting in substantial reductions in curation costs. Third, detailed knowledge about the subsurface characteristics of a site can dramatically reduce the often prohibitive excavation costs while increasing the positive results. The absence of this knowledge also can lead to the unintended disturbance of human remains and other culturally sensitive deposits during subsequent construction or military training, requiring costly and complex action to comply with the National American Graves Protection and Repatriation Act. (Project Completion - 2006)