Comparison of source-location algorithms for atmospheric samplers.

Area of Science:

• Atmospheric radionuclide transport modeling
• Nuclear monitoring and verification technologies

Background:

• Numerous algorithms exist for determining atmospheric radionuclide release source characteristics.
• The International Monitoring System (IMS) collects data for the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO).

Purpose of the Study:

• Compare the performance of three source-location models for radionuclide releases.
• Evaluate model accuracy in estimating source event parameters using IMS data.

Main Methods:

• Compared a deterministic Possible Source Region (PSR) model with two Bayesian models (BAYEST and FREAR).
• Evaluated models using 100 synthetic release cases of the xenon isotope ¹³³Xe.
• Utilized a fictitious network of 120 noble gas samplers with varying collection periods (3-24 hours).

Main Results:

• Bayesian models (BAYEST, FREAR) provided more accurate location estimates than the deterministic PSR model.
• FREAR showed slightly better location performance than BAYEST.
• Location accuracy improved with reduced sample collection times for BAYEST.
• Bayesian models estimated release magnitude, unlike the base PSR model.
• Poor location/time estimates occurred when the release point was distant from samplers, irrespective of model.

Conclusions:

• Bayesian source-location algorithms offer superior accuracy and magnitude estimation capabilities compared to the PSR approach.
• Sampling strategy, particularly collection duration, impacts estimation accuracy.
• Even dense sampling networks can produce unreliable results under specific release scenarios.