It is estimated that more than 11 million acres of land in the United States potentially contain unexploded ordnance (UXO). The UXO characterization problem is primarily associated with current and former military testing ranges, of which Department of Defense (DoD) sites in the process of being returned to the public sector are particularly critical. These DoD sites include bases designated for the base realignment and closure process and formerly used defense sites. In many cases, actual UXO contamination occurs in only small portions of a site under investigation. The large amount of potentially contaminated land area and the observation that the actual UXO contamination may occur in only a fraction of this area make efficient characterization for the presence of UXO extremely difficult. In particular, the potential for unnecessary expenditure of characterization resources is high.

The objective of this project was to develop a sampling protocol for estimating the intensity of UXO contamination across a site. The protocol uses an inherently Bayesian approach that allows for the incorporation of historical information and geophysical data into the site characterization process.

Estimates of the total number of anomalies within each model cell.

This 2-year project integrated the following four distinct tasks: (1) development of a theoretical framework for the characterization of UXO sites based on Bayesian statistics and interpretation of the spatial distribution of UXO as a spatial Poisson process; (2) development of a sampling protocol for practical field characterization of UXO sites; (3) implementation of data worth techniques to direct the specific location and number of detailed geophysical surveys that are necessary; and (4) validation of the procedures developed herein against a test data set.

Controlled testing of the site characterization approach was completed on simulated data sets created by Mitretek. Researchers demonstrated the flexibility of the site characterization approach to create spatial estimates of different attributes that might be measured on a site and the use of probability mapping to define the edges of targets and to locate additional sampling transects. Four example calculations using both a simulated site and actual data collected at the Pueblo of Laguna N-11 site in New Mexico were completed. These calculations provide a wide range of examples in which geostatistical methods can be applied to UXO site characterization problems. In addition to these example applications, a theoretical limitation of the basic geostatistical estimation algorithm, kriging, caused by redundant data as collected along transects was investigated. The site characterization protocol is now being demonstrated under ESTCP project MR-200325.

This project provides a new UXO characterization protocol that can be applied to DoD sites. The protocol improves on currently available techniques used to characterize UXO sites by incorporating prior information through a Bayesian approach, using geostatistical techniques to update that prior information, and optimizing characterization using a data worth approach. (Project Completed – 2003)