Quantitative visualization of a radioactive plume using gamma-ray imaging spectrometry: estimation of initial release condition by machine learning and feasibility of real-time analysis in emergency monitoring applications

A novel monitoring method for the quantitative visualization of 3D distribution of a radioactive plume accidentally released from a nuclear facility was proposed, and the feasibility of its analysis method was demonstrated in our previous study by means of a preliminary test using hypothetical data. The method combines gamma-ray imaging spectroscopy with an Electron Tracking Compton Camera (ETCC) and real-time high-resolution atmospheric dispersion simulation driven by wind observations. The 3D plume distribution is reconstructed inversely from the direct gamma-ray images by several ETCCs deployed around the target area. This reconstruction employs a regularized optimization in which the air‑concentration field predicted by the dispersion simulation serves as prior information. We also prototyped a method to identify a release point, which is required as input conditions for the atmospheric dispersion calculation, solely from gamma-ray images of the ETCC immediately after the release. A method based on machine learning was employed in test analyses targeting the Fukushima Daiichi Nuclear Power Station. The results show good performance in identifying the release point when using theoretical gamma-ray image data. In practical aspect, the scope of application and accuracy of the method was evaluated using hypothetical gamma-ray image data under realistic assumptions. Moreover, the feasibility of real-time analysis was confirmed by implementing an analysis workflow designed based on the 10-minute interval data acquisition.