A Distribution-Free Approach for Detecting Radioxenon Anomalous Concentrations

The detection of anomalous radioxenon atmospheric concentrations plays a key role for revealing both underground nuclear explosions and radioactive emissions from nuclear power plants and medical isotope production facilities. For this purpose, the method currently used by the International Data Center of the CTBTO is based on descriptive thresholds. In this work we propose a statistical inference-based method, that allows to distinguish between the typical background of atmospheric radioxenon and anomalous values above background. We used a non-parametric methodology that does not require any assumption on the phenomenon distribution. In such a way we overcome the problem due to the non-normality of the radioxenon data.