By Johan Beekhuizen
SUMMARY
To prepare for the unlikely event of a nuclear emergency, atmospheric dispersion models are used to assess possible release concentrations and radionuclide deposition rates. In the Netherlands, the NPK-PUFF dispersion model is used to provide critical information about the radioactive plume during the first phase of a nuclear accident. By combining NPK-PUFF model predictions with spatial analyses of real-time gamma (γ) dose rate measurements from the Dutch National Radioactivity Monitoring Network (153 stations and 14 α/β air sampling monitors), informed intervention decisions can be made. In an emergency, additional mobile devices can be deployed to increase the accuracy in spatial estimates of radiation exposure. The main objective of this research was to determine where to optimally locate mobile devices in order to minimize costs associated with uncertain estimates of the radioactive plume. A Monte Carlo approach was used to account for uncertainty in the NPK-PUFF model. First probability distribution functions (pdfs) of model input parameters that contribute most to uncertainty in the predictions were investigated. Next many possible developments of these input parameters could be simulated and used to create various model outputs. To account for uncertainty in the model itself, a stochastic residual was generated and added to the NPK-PUFF model output. This resulted in a set of simulated realities. The optimal location of mobile measuring devices was then determined using spatial simulated annealing. The optimisation criterion selected for this research was the cost of making incorrect decisions due to prediction uncertainty. This uncertainty was quantified by comparing predicted maps with simulated realities. The predicted maps were created by adding a residual to a reference dose map. The residual was calculated by interpolating the differences between observations of the simulated reality and reference dose map using Inverse Distance Weighting. The cost function took population density into account and permitted a weighting of false positive (predicted concentrations greater than intervention thresholds when the actual value is below) and false negative values. Results indicate that optimal placement of mobile devices in the event of an accident tend to be in areas at the edge of the advancing radioactive plume, in densely populated areas. However, the Inverse Distance Weighting interpolation gave unsatisfactory results in improving the predicted maps.
Keywords: Emergency; Uncertainty; Meteorological predictions; Radioactivity; Sampling design; Spatial simulated annealing