Mapping distributed radiation sources is vital for radiological emergency response, for example, in reactor accidents such as Chernobyl or Fukushima. In particular, quantitative maps of radioactivity concentrations can inform path-planning methods for keeping the dose to first responders below established thresholds. The task of radiation mapping, quantitatively, is challenging and requires a knowledge of absolute detector efficiencies, accurate models of the 3D scene being mapped, and quantitatively-correct image reconstruction algorithms. Moreover, producing distributed radiological sources with known ground truth for developing the techniques is challenging from an operational perspective. Distributing radioactive material in a liquid, powder, or aerosol form can be difficult to do in a configurable, repeatable fashion, and can present a substantial human and environmental safety hazard.
To overcome these concerns, scientists from LBNL’s Applied Nuclear Physics program developed a method for emulating true distributed sources with arrays of sealed point sources. This technique avoids the risk of inhaling or ingesting radioactive material, and allows for easier ground-truthing and reconfiguration of source patterns during measurements. The Berkeley Lab team partnered with Washington State University and Idaho National Laboratory to irradiate copper pellets in the WSU nuclear reactor to produce hundreds of sealed ~5 mCi Cu-64 sources. The sources were arranged on a field in various patterns, each covering approximately 100 square meters, and radiation detectors were then flown over the source patterns on an unmanned aerial system (UAS)—see Figure 1.
Analysis is ongoing to compare the measured results against the known ground truth source distributions. Preliminary results show that the shapes of the source distributions can be accurately reconstructed, and that the overall intensity can be determined to within the systematic uncertainties of the measurements. The measurements will also form the basis for radiation mapping uncertainty quantification studies, with a focus on accurate methods that can be computed in near-real time on edge computers, and will stimulate future work on training UAS swarms to efficiently and autonomously map sources during radiological emergencies.
Jayson Vavrek led the experimental design and analysis team for this work. Jayson, along with Brian Quiter, Mark Bandstra, and Daniel Hellfeld collected the experimental data in August 2021. Tenzing Joshi is PI of the project supporting the work.