Bayesian source reconstruction of an anomalous Selenium-75 release at a nuclear research institute.

Abstract Atmospheric transport and dispersion models are important tools in radiation protection as they help to estimate the impact of radionuclides released into the atmosphere. In particular, such models can be used in combination with radionuclide observations to estimate unknown source term parameters, or to improve source term estimates obtained through other methods. In this paper, a Bayesian inference system was used to determine the source term parameters and their corresponding credible intervals of a real-world anomalous 75Se release at a nuclear facility in Belgium. Furthermore, a formulation is proposed that not only takes into account true detections, but also true instrumental non-detections, false alarms and real misses. The Bayesian inference system is able to correctly determine the known source location. The Bayesian inference is then refined by fixing the release location and by making stronger assumptions about the release period.