The impact of below detection limit samples in residual risk assessments for decommissioning nuclear power plant sites

Abstract

As the number of nuclear power plants (NPPs) being decommissioned increases internationally, many issues are being raised. One such issue is related to site soil analyses for the determination of residual risk for license termination. In a typical site-cleanup analysis, the majority of soil samples at the site are at or below the detection limit (BDL). Conventional approaches to managing BDL data are to simply ignore or substitute the data with a value of zero or the detection limit itself. However, these approaches are statistically biased, limiting their usefulness. Within the environmental science community, the issue of how to treat BDL data has been examined by a number of investigators. This study reviewed the issue of BDL data in nuclear decommissioning using the analytical methods suggested by studies in the environmental science, including the Kaplan-Meier method, robust regression on order statistics, and maximum likelihood estimation. The use of these methods to handle BDL data was examined using a case study with respect to its potential impact on dose/risk assessment, the soil volume removal estimate, and the associated costs. The case study was based on the Colorado School of Mines Research Institute's site soil data. Our analysis included the consideration of the uncertainties associated with residual dose/risk, waste soil volume estimation, and costs. An uncertainty analysis was conducted using a Latin hypercube sampling approach. Results showed that using BDL data can have a large impact on the estimation of dose/risk, waste volume, and waste disposal cost of a NPP decommissioning project.