AUTOMORC – Improvement of automatic methods for identification of radioactive material out of regulatory control (MORC) by mobile gamma spectrometric search experiments. Report 2017

Activity Acronym:

AUTOMORC

Authors:

Christopher L. Rääf (chair), Robert R. Finck (co-chair), Antanas Bukartas, Gísli Jónsson, Sune Juul Krogh, Simon Karlsson, Morten Sickel, Petri Smolander, Jonas Wallin, Robin Watson,

Abstract:

A model for calculating maximum detection distances in mobile search for lost gamma radiation sources was developed in a previous NKS supported project MOMORC 2016. A first validation of the model's correctness was done in a field experiment in September 2016 with participants from all Nordic countries. The model showed that maximum detection distances could be predicted within ± 30 m, which also was the limitation of the experiment. The model also showed that using combined analysis of sets of measurements along a vehicle path could be used to detect a possible source. The count rate from primary photons in the detector when driving past a point source at constant speed is depending on the source activity and the distance between the road and the source. The count rate versus time (the “intensity” curve) depends only on the distance to the source when the speed is fixed and the vehicle path is straight. This fact can be used to determine the distance to a source from a set of measurements. The problem can be solved with Bayesian statistical methods.
A Bayesian based Markov chain Monte Carlo method was used to determine the location and activity of some of the point sources in the MOMORC 2016 experiment. The method seems to produce reasonably correct values of distances to sources, provided that the count rate statistics in the detector is high enough for the computational algorithm to identify the “signal” (as an “intensity curve). The statistical uncertainty in the distance determination using measurement data from a 128 % HPGe-spectrometer was in the order of ±20%. The uncertainty in the activity determination was larger, especially for NaI(Tl)-spectrometers. This could be due to a systematic deviation in the efficiency calibration of the detectors, because when testing the method with synthetic data the activity calculations gave reasonably correct results.
The position of a source can be shown on a map as likelihood for a source present at a certain location together with its activity coupled to that location. Five examples of calculated likelihood locations based on Bayesian analysis of mobile measurements are given in the report.

Keywords:

Mobile gamma spectrometry, orphan source search, detection distance, Bayesian analysis, Markov chain Monte Carlo