Many formerly used defense sites (FUDS) are slated for cleanup and transfer to the public for other uses. The risks associated with unexploded ordnance (UXO) in soils at these sites are a significant concern and must be managed to acceptable levels commensurate with hazardous cleanup activities and intended future use. Risks include the failure to identify UXO that could be discovered inadvertently after cleanup and spending too many tax dollars on non-optimal characterization schemes. Because many of the FUDS involve large geographical areas, it is often impractical and cost prohibitive to characterize the entire site. Defensible statistical sampling methods are needed to balance the characterization requirements against the cost and feasibility constraints.

The objective of this project was to evaluate statistical methods and tools for use in developing characterization and verification plans and data evaluation schemes. Development of statistical sampling methods and tools was consistent with the Environmental Protection Agency’s (EPA) Data Quality Objectives (DQO) process, which is used to plan any characterization activity to ensure that the right type, quantity, and quality of data are gathered to support confident decision making.

The methods developed strike a balance between the probability of missing UXO and the costs of unnecessarily burdensome characterization. Statistical methods were evaluated and adapted or developed; prototype tools were developed and demonstrated. The methods allow quick evaluations of tradeoffs involving costs, the risk of missing UXO, acceptable probabilities for decision errors, the percentage of the site characterized, false positive rates, and grid sizes. Defensible statistical sampling schemes can optimize the amount of characterization required for areas where UXO is highly unlikely.

This project developed methods and tools to aid in finding unknown high-likelihood UXO areas and providing confidence that large unsurveyed areas are unlikely to contain UXO. Additional work examined statistical approaches to evaluating whether individual anomalies on a dig list were likely UXO. This methodology can be used in determining when it is acceptable to conclude that continued digging at locations on the dig list is unlikely to uncover any additional UXO. The target area detection and compliance methods developed have been incorporated into the Visual Sample Plan software code so non-statisticians can implement the survey design methods in a systematic planning DQO framework. Validation and demonstration efforts for the target area detection objectives conducted to date under ESTCP project MM-0325 indicate that the survey design methods can indeed be used effectively to find target areas that are likely to contain UXO.

There are numerous sites where UXO is a significant concern, many of which are very large with varied terrain. Complete characterization of these sites is not feasible with existing technology and cleanup budgets. Developing defensible sampling schemes for bounding UXO contamination areas is an important factor in reducing the cost burden. (Project Completed – 2004)