Data assimilation for assessing the deposition of radio-nuclides after a nuclear accident

Full text: Model predictions of possible radiological consequences after an accidental release of radionuclides play an important role in nuclear emergency management. Amongst other information, predictions of the deposition to the ground are of high importance. This poster describes how model predictions of the deposition can be improved by using monitoring data. This process of combining model predictions and observations is usually referred to as data assimilation. Input data to the deposition modelling are mainly results of an atmospheric dispersion model like concentration of radionuclides in air and rain water. From this data the model calculates the activity deposited on pasture, lawn and up to 22 plant types. The model results are associated with considerable uncertainties, due to uncertainty in model parameters but especially due to the uncertainty of input data from dispersion modelling. Several factors determine this uncertainty: uncertainty in the input data about the amount of activity released, its nuclide composition, the exact release times, the effective release height, the wind field etc.; uncertainty of model parameters like dispersion parameters and plume rise parameters; model limitations, e.g. for dispersion in complex terrain; stochastic nature of turbulent dispersion. The objective of introducing data assimilation into the deposition modelling is to improve the predicted deposition and reduce its uncertainty by making use of measurements of gamma dose rates after the deposition process has ended. Data assimilation promises to enhance the quality of deposition predictions, since the model is used to spread the measurement information both in space and time: even few measurements at any locations and any time can improve the prediction; at any time, all measurements from the past are used to optimize model predictions of the time-dependence of ground contamination; measurements at individual locations can improve the predictions at the whole calculation grid; dose rate measurements above one type of vegetation also allow to update the contamination of other types of vegetation; total dose rate measurements allow updating in parallel the ground contamination for different radionuclides. Data assimilation in the deposition modelling is performed with a Kalman filter approach, which is especially tailored towards operational use. Essential for data assimilation with the Kalman filter is the correct propagation of uncertainties throughout the model. The data combination scheme of the Kalman filter considers the relative uncertainties of both model predictions and measurements. Furthermore, the uncertainty propagation allows to provide decision makers with uncertainty estimates for all results. Similar data assimilation capabilities are also under development for other models for off-site nuclear emergency management within the framework of the DAONEM project. fig. 1 (author)