1 Department of Informatics and Mathematical Modeling, Technical University of Denmark2 Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark3 Department of Management Engineering, Technical University of Denmark4 Department of Applied Mathematics and Computer Science, Technical University of Denmark
During a nuclear accident in which radionuclides are released to the atmosphere, off-site dose assessment using atmospheric dispersion models play an important role in facilitating optimized interventions, i.e. for mitigating the radiological consequences. By using data assimilation methods, radiological observations, e.g. dose rate measurements, can be used to improve these model predictions and to obtain real-time estimates of the atmospheric dispersion parameters. This thesis examines data assimilation in the context of atmospheric dispersion of radioactive materials. In particular, it presents a new method for on-line estimation of the radionuclide source term, i.e. the amount and composition of the released radionuclides, and the main dispersion parameters, based on radiation monitoring data obtained in the vicinity of the release. The method is based on the extended Kalman filter (EKF) and a stochastic nonlinear state space model formulation, where the state variables are the significant model parameters. The atmospheric dispersion and the ensuing radiation field are modeled by a static Gaussian plume model, which is applicable up to a few kilometers from the release. The embedded parameters of the state space model are determined by maximum likelihood estimation, making the approach essentially free of external parameters. The proposed Kalman filter method is tested against both simulated data as well as real radiation monitoring data from a new atmospheric dispersion experiment. For a circular array of gamma detectors, the method may be used for on-line estimation of the source term, the release height of the radionuclides and the main plume advection direction. These parameters may in turn be used as input to long- and short-range atmospheric dispersion models, resulting in greatly improved dose rate assessment. The Kalman filter method is found to be computationally efficient and therefore has the potential of becoming an efficient operational assimilation tool for nuclear emergency management. Empirical dispersion data are crucial in order to evaluate dose rate models and data assimilation methods in a realistic setting. New experimental studies of atmospheric dispersion of radioactive material was carried out in October 2001 at the SCK"CEN in Mol, Belgium. In the Mol experiment, the radiation field from routine releases of 41 Ar is recorded by an array of gamma detectors along with simultaneous measurements of the 41 Ar source term, the main meteorological parameters and direct plume measurements, using a Lidar scanning technique. The thesis provides a detailed description of the experiment and the subsequent data analysis. A reference dataset has been generated, which is suitable for evaluation of gamma dose rate models and for development and testing of data assimilation methods for atmospheric dispersion of radioactive materials. The Mol dataset has been used for experimental evaluation of the Gaussian plume model and the RIMPUFF model; the results of these studies are presented in the thesis.