Model uncertainties in bayesian source localisation using the inverse modelling tool frear

A global network of monitoring stations is set up that can measure tiny concentrations of airborne radioactivity as part of the verification regime of the Comprehensive Nuclear-Test-Ban Treaty. If Treaty-relevant detections are made, inverse atmospheric transport modelling is one of the methods that can be used to determine the source of radioactivity. In order to facilitate the testing of novel developments in inverse modelling, two sets of test cases have been constructed using real-world ¹³³Xe detections associated with routine releases from a former medical isotope production facility. One set consists of 24 cases with 5 days of observations in each case, and another set consists of 8 cases with 15 days of observations in each case. A series of inverse modelling techniques and several sensitivity experiments have been applied to determine the (known) location of the medical isotope production facility. Metrics have been proposed to quantify the quality of the source localisation. From that, it has been found that the Bayesian inference underestimated uncertainties. Here, two approaches are explored to address the issue of underestimating the uncertainty in Bayesian source reconstruction.