Optimizing Sensor Data Interpretation via Hybrid Parametric Bootstrapping

TL;DR

The paper tackles estimating upper bounds for $^{235}$U concentrations at nuclear legacy sites using small environmental datasets. It proposes a hybrid parametric bootstrap (HPB) combined with Steiner's Most Frequent Value (MFV) to handle outliers and per-sample uncertainties, and it compares against nonparametric bootstrap and Chebyshev-based bounds. The authors apply these methods to a public NORM dataset from Egypt's Eastern Desert, demonstrating that HPB yields robust, often conservative upper limits suitable for remediation planning and regulatory compliance. Overall, the work provides a practical, data-efficient toolkit for criticality-safety assessments under data scarcity at legacy nuclear facilities.

Abstract

The Chalk River Laboratories (CRL) site in Ontario, Canada, has long been a hub for nuclear research, which has resulted in the accumulation of legacy nuclear waste, including radioactive materials such as uranium, plutonium, and other radionuclides. Effective management of this legacy requires precise contamination and risk assessments, with a particular focus on the concentration levels of fissile materials such as U-235. These assessments are essential for maintaining nuclear criticality safety. This study estimates the upper bounds of U-235 concentrations. We investigated the use of a hybrid parametric bootstrapping method and robust statistical techniques to analyze datasets with outliers, then compared these outcomes with those derived from nonparametric bootstrapping. This study underscores the significance of measuring U-235 for ensuring safety, conducting environmental monitoring, and adhering to regulatory compliance requirements at nuclear legacy sites. We used publicly accessible U-235 data from the Eastern Desert of Egypt to demonstrate the application of these statistical methods to small datasets, providing reliable upper limit estimates that are vital for remediation and decommissioning efforts. This method seeks to enhance the interpretation of sensor data, ultimately supporting safer nuclear waste management practices at legacy sites such as CRL.

Figures (6)

• Figure S1: A conceptual site model for a nuclear waste area at Chalk River Laboratories.
• Figure S2: Visualization of $^{235}\text{U}$ specific activity: median, IQR, and outliers (red dots on the plot).
• Figure S3: Histogram showing the $^{235}$U datasets, including the original dataset with outliers (bottom) and the dataset without outliers (top). The values are listed in the text. In each histogram, a thin solid vertical line represents the MFV for the entire dataset, and a thin dashed line represents the mean value.
• Figure S4: Histograms showing the $^{235}$U NORM dataset without outliers (26 elements, top) and a smaller subset (9 elements, bottom). The corresponding values for these datasets are provided in the text.
• Figure S5: Histogram showing the MFV from traditional nonparametric bootstrap samples of the $^{235}$U NORM dataset without outliers (26 elements). In the histogram, a thin vertical solid line represents the MFV (2.19), whereas thin dashed lines represent the lower (2.07) and upper (2.30) 95.45% confidence limits.