Abstract
Radioactive emissions caused by severe accidents pose a great threat to the environment and creatures along the spread. The source inversion technique is designed to retrieve the corresponding release rate based on environmental monitoring data and has been widely integrated into the nuclear accident consequence assessment. The Model error and its propagation in the inversion cause the estimate unrealistic and confuse the real releases. Several self-correcting source inversion methods have been proposed to handle it, for instance, the joint estimation method that simultaneously retrieves the estimate and corrects the model bias by introducing a correction coefficient matrix. In this study, the joint estimation method has been investigated against a wind tunnel experiment with the incoming airflow from the southeast direction (SSE). Three nuclear power plants were located in the scenario, surrounded by mountains and sea. The Risø Mesoscale Lagrangian PUFF model (RIMPUFF) was used for atmospheric diffusion predictions and the transport matrix construction. The fine wind tunnel experimental data was used as the observations to drive the standard (Tikhonov) inversion method and the joint estimation method. The results show that the release rate estimated by the Tikhonov method was four times higher than the ground truth on the representative SSE direction. In comparison, the self-correcting joint estimation method reduces the error to only 4%.
