The problem of evaluation of radiation monitoring network quality is considered. The comparison of possible network configurations is performed on simulated experiments where a twin model is used to simulate observed data. An assimilation procedure based on the particle filter approach is run for each configuration of the network. Specifically, multiple realisations of segmented Gaussian plume models are computed, each with different parameters. The simulated measurements from the network sensors are used to weight the likelihood of the parameter realisations, providing an empirical posterior distribution of the parameters. The quality of the network is then evaluated as an expected value of multiple loss functions with respect to the empirical posterior distribution. Our goal is to provide a tool that allows the decision-makers to compare the proposed configurations from various points of view. The results for a hypothetical discharge of radioactivity are presented.
Keywords: dispersion modelling, dose rates, assimilation, monitoring network