Enhancing Bayesian methods for radioactive source localization: a parameter study, prior construction and signal...

Area of Science:

• Nuclear physics and instrumentation
• Computational physics and data analysis
• Radiation detection and measurement

Background:

• Direct source localization is often infeasible in hazardous environments or for low signal-to-noise ratio (SNR) scenarios.
• Bayesian analysis offers a probabilistic approach to parameter estimation and uncertainty quantification.
• PyMC is a powerful Python library for probabilistic programming.

Purpose of the Study:

• To clarify the practical advantages of Bayesian analysis for gamma-ray source localization.
• To present a PyMC-based Bayesian framework for localizing unshielded gamma-emitting sources and estimating their activity.
• To evaluate the performance of generic versus informed priors and assess the impact of signal-to-noise ratio (SNR) enhancement techniques.

Main Methods:

• Development of a PyMC-based Bayesian framework for source localization and activity estimation.
• Utilization of both generic and measurement-derived informed priors.
• Evaluation using 1240 synthetic datasets with varying source parameters and background levels.
• Testing with controlled field experiments and application of Savitzky-Golay smoothing.

Main Results:

• Informed priors demonstrated improved parameter stability near detection limits and reduced computational time compared to generic priors.
• Savitzky-Golay smoothing enhanced SNR and robustness in marginal detection cases, but could not overcome insufficient signal strength.
• The two-step approach using informed priors effectively constrained the search area and improved analysis efficiency.
• Bayesian localization proved operationally beneficial under specific conditions, particularly for large-area searches and low SNR environments.

Conclusions:

• Bayesian localization using informed priors is operationally beneficial for gamma-ray source detection in challenging environments.
• The PyMC framework provides a flexible and effective tool for probabilistic source localization.
• Careful consideration of prior selection and SNR enhancement is crucial for optimal performance in low-signal regimes.