Source Localization of Accidental Radionuclide Release Using a Data Assimilation Method Based on Forward and Backward Dispersion Simulations

Source localization of accidental radionuclide release is crucial to nuclear emergency and decision making, especially when no reliable information about possible release regions can be previously obtained. To rapidly and accurately determine the source location with a limited number of measurements, this paper proposes a two-step data assimilation method based on forward and backward dispersion simulations in the RIMPUFF model. In the first step, meteorological variables such as wind speed are reversed and the measurements are set as sources for several backward RIMPUFF runs. By identifying the overlap of backward plumes originating from different measurement sites, potential regions discretized into grids can be roughly determined. In the second step, the grids selected by the first step are set as sources for forward RIMPUFF runs with unit release rate and the source location will be refined by minimizing a correlation-based function of simulated and measured data. The method is verified by SCK-CEN 41Ar field experiment. Using gamma dose rate data from all measurement sites, source location is retrieved with errors of 22.36m and 30m respectively on the two days of the field experiment. Compared with the direct correlation-based method, the proposed method achieves satisfactory solutions in an obviously shorter time. In particular, there is no need for predetermining the source release rate throughout the process, indicating that the method can conveniently combine with other source term inversion approaches whose source location is viewed as a known quantity. Therefore, the presented two-step data assimilation method here is potentially a useful tool with high accuracy and efficiency for further integration in nuclear emergency response system.