Monitoring airborne radioactivity involves determining the temporal profile and source characteristics of the release. While the location of the release is often known, the temporal profile and total quantity of the released substances are typically uncertain or only partialy known. This research focuses on reconstructing the temporal dynamics of a release using field measurements, such as concentration or deposition data. The methodology employs optimization techniques to compare observed measurements with numerical predictions from atmospheric dispersion models. Advanced Bayesian inference methods in machine learning, along with state-of-the-art deep neural networks, are utilized to enhance this process. Primary applications include the characterization of radionuclide releases from specific locations or estimation of complex spatial-temporal sources, such as emissions of ammonia or microplastics and microfibers.

FIND OUT MORE