Abstract:Abstract In recent years, Europe has been confronted with a series of radionuclide leak incidents, emanating from sources that remain unidentified. Such incidents are typically assumed to originate from a single, static point source, allowing the source location to be pinpointed to a specific, time-unvarying coordinate. However, the unavailability of detailed information about these leaks, including crucial data like the operating status of involved nuclear plant, renders direct estimation of release details using a top-down approach unfeasible. Consequently, a bottom-up approach, relying on environmental measurements and atmospheric dispersion modeling, becomes the only feasible choice. This approach focuses on the spatiotemporal source reconstruction of atmospheric radionuclide release, aiming to optimize the source location and release rate by minimizing the discrepancy between simulations and measurements. Owing to the interdependence of these spatiotemporal source parameters, traditional source reconstruction methods often attempt to resolve both temporal and spatial aspects simultaneously. The resulting tremendous solution space may cause substantial estimation errors in both the source location and release rate. To tackle this issue, this paper proposes a novel framework for spatiotemporally decoupled source reconstruction. This framework splits the reconstruction into the source localization and release rate estimation stage, reducing the solution space at each stage. Initially, an artificial neural network model is employed to map a relationship between the source location and several simulation-measurement error metrics. The error metrics for each source location are obtained based on the release rates calculating by the Tikhonov regularization method. Subsequently, a likelihood function is constructed based on these error metrics, followed by assessing the posterior distribution of the source location using the Bayesian statistics. Finally, based on the estimated source location, the release rate is calculated using the Tikhonov regularization method. The proposed framework has been first validated using a European Tracer Experiment (ETEX) whose source is previously known. The source location error was less than 100 km and the estimated release rate profile match the truth well. Furthermore, this framework was applied to reconstruct the source of the 2017 106Ru leak incident, a case where the source remains unknown. The estimated source location and release amount are consistent with public literature. The successful application of this framework to both controlled experiment and real-world radionuclide leak incident underscores its potential as a key component of the nuclear emergency response system.

Comparing machine learning and inverse modeling approaches for the source term estimation

Automated release rate inversion and plume bias correction for atmospheric radionuclide leaks: A robust and general remediation to imperfect radionuclide transport modeling

An inverse method to estimate emission rates based on nonlinear least-squares-based ensemble four-dimensional variational data assimilation with local air concentration measurements

Comparison Between Different Source Reconstruction Methods of Atmospheric Leakages

Comparison of source-location algorithms for atmospheric samplers

Coupled Source Term Estimation Using Both Forward and Backward Method for Nuclear Power Plant Accident

Source term estimation for continuous plume dispersion in Fusion Field Trial-07: Bayesian inference probability adjoint inverse method

Inverse calculation approaches for source determination in hazardous chemical releases

Source Term Estimation in Nuclear Accidents Using 4d Variational Data Assimilation for Heterologous Atmospheric Condition

Source term inversion coupling Kernel Principal Component Analysis, Whale Optimization Algorithm, and Backpropagation Neural Networks (KPCA-WOA-BPNN) for complex dispersion scenarios

Identifying Atmospheric Pollutant Sources Using a Machine Learning Dispersion Model and Markov Chain Monte Carlo Methods

A Spatiotemporally Separated Framework for Reconstructing the Sources of Atmospheric Radionuclide Releases

Source Localization of Accidental Radionuclide Release Using a Data Assimilation Method Based on Forward and Backward Dispersion Simulations

Inversion Method for Multiple Nuclide Source Terms in Nuclear Accidents Based on Deep Learning Fusion Model

A Novel Framework for Spatiotemporally Decoupled Source Reconstruction of Radionuclides Released into the Atmosphere

Source term inversion of nuclear accidents based on ISAO-SAELM model

Towards new simplified methodologies on source term estimation and associated uncertainties from accidental airborne releases

Simultaneous Release Rate Estimation and Modeled Plume Bias Correction for Atmospheric Radionuclide Emissions

Bayesian hybrid-kernel machine-learning-assisted sensitivity analysis and sensitivity-relevant inverse modeling for groundwater DNAPL contamination

Source term inversion of short-lived nuclides in complex nuclear accidents based on machine learning using off-site gamma dose rate

Nuclear accident source term estimation using Kernel Principal Component Analysis, Particle Swarm Optimization, and Backpropagation Neural Networks