In the Fukushima nuclear accident, due to the lack of field observations and the complexity of source terms, researchers failed to estimate the source term accurately immediately. Data assimilation methods to estimate source terms have many good features: they works well with highly nonlinear dynamic models, no linearization in the evolution of error statistics, etc. This study built a data assimilation system using the ensemble Kalman Filter for real-time estimates of source parameters. The assimilation system uses a Gaussian puff model as the atmospheric dispersion model, assimilating forward with the observation data.
Considering measurement error, numerical experiments were carried on to verify the stability and accuracy of the scheme. Then the sensitivity of observation configration is tested by the twin experiments.
First, the single parameter release rate of the source term is estimated by different sensor grid configurations. In a sparse sensors grid, the error of estimation is about 10%, and in a 11*11 grid configuration, the error is less than 1%.
Under the analysis of the Fukushima nuclear accident, ahead for the actual situation, four parameters are estimated at the same time, by 2*2 to 11*11 grid configurations. The studies showed that the radionuclides plume should cover as many sensors as possible, which will lead a to successful estimation.
Uncertainty study on atmospheric dispersion simulations using meteorological ensembles with a Monte Carlo approach, applied to the Fukushima nuclear accident
