Assessment of mobile radiometry data in radiological emergencies using Bayesian statistical methods

Abstract: Nuclear technology and the use of radiation sources have been extensively adopted in the modern world. Despite that, a number of radiological accidents, significantly affecting population and the environment, has happened. Few of the most known are accidents in Chernobyl and Fukushima nuclear power plants. As a result, significant areas are still contaminated with some residential areas evacuated. To estimate the possibility to return to an affected area, the spatial distribution of deposition density of the radionuclides within the area has to be evaluated using mobile gamma spectrometry. In radiological emergencies involving lost radioactive sources, mobile gamma spectrometry equipment is used to survey the area where the source is suspected to be, trying to localize and identify the radioactive source. For both: radionuclide spatial distribution estimation and source localization problems, the most prominent methods of data analysis were established decades ago and are robust and easy to utilise. Despite that, the usual methods, like interpolation or using an alarm threshold level, do not use all of the available data in the measurement time‑series to produce an estimation of the situation.Using Bayesian statistics it is possible to combine the prior knowledge about the situation with the data, to obtain predictions about the situation. Thus, the aim of this thesis was to investigate the feasibility of a Bayesian‑based approach for mobile gamma spectrometry applications in radiological emergencies. A Bayesian algorithm was developed for analysing measurement time-series to extract the physical location and source strength in an orphan source search and to map radionuclide deposition in an area of interest. It was found that the Bayesian methods could be successfully applied to obtain the predicted position and activity of the source or the spatial deposition of the radionuclide within an area. It was also found that the accuracy of the Bayesian estimations is heavily dependent on the quality and quantity of the data, the more and the better quality data – the better the assessments. A comparison of the most prominent method for an orphan source detection using alarm threshold levels with the Bayesian method, in terms of detection probability of orphan sources using simulated data, has shown that the Bayesian algorithm can potentially detect the radioactive sources more reliably than the usual alarm threshold method due to the inclusion of all of the data in the measurement time-series.The overall conclusion of the thesis is that Bayesian methods can be successfully applied to mobile spectrometry data. The mapping and positioning of gamma emitting radionuclides can be done more precisely and provide more information about the radiation in the environment during radiological accident scenarios. Despite the shown potential of Bayesian methods in mobile gamma spectrometry within this thesis, further investigations are needed to validate the findings discussed.