In the early phase of a nuclear accident two large sources of uncertainty exist: one related to the source term and one associated with the meteorological data. In brief, operational methods will be developed in AVESOME for quantitative estimation of uncertainties in atmospheric dispersion modelling resulting from uncertainties in assessments of both the release of radionuclides from the accident and their atmospheric dispersion.
Previously, due to lack of computer power, such methods could not be applied to operational real-time decision support. However, with modern supercomputing facilities, available e.g. at national meteorological services, the proposed methodology is feasible for real-time use, thereby adding value to decision support.
In the recent NKS-B projects MUD and FAUNA as well as the ongoing project MESO, the implications of meteorological uncertainties for nuclear emergency preparedness and management have been studied, and means for operational real-time assessment of the uncertainties in a decision-support system (DSS) have been described and demonstrated.
In the proposed project, AVESOME, we address the uncertainty of the radionuclide source term, i.e. the amount of radionuclides released and the temporal evolution of the release. Furthermore, the combined uncertainty in atmospheric dispersion model forecasting stemming from the source term and the meteorological data is examined. Ways to implement the uncertainties of forecasting in DSSs, and the impacts on real-time emergency management will be described.
The proposed methodology will allow for efficient real-time calculations by making use of scaling properties in the equations governing the release and the atmospheric dispersion of radionuclides. Accordingly, the computer-resource demanding calculations should be carried out at the high-performance computing facilities available e.g. at the national meteorological services, whereas less demanding post-processing could be carried out at the computer hosting the DSS. The former tasks include the atmospheric dispersion model calculations and Monte Carlo sampling of the release and atmospheric dispersion model data, the latter includes interactive communication with the supercomputer as well as presentation of final results; either in the form of probability distributions or in the form of scenario-based radionuclide concentrations and depositions.