Model Intercomparison of Atmospheric 137Cs From the Fukushima Daiichi Nuclear Power Plant Accident: Simulations Based on Identical Input Data

A model intercomparison of the atmospheric dispersion of cesium‐137 (137Cs) emitted after the Fukushima Daiichi Nuclear Power Plant accident in Japan was conducted to understand the behavior of atmospheric 137Cs in greater detail. The same meteorological data with a fine spatiotemporal resolution and an emission inventory were applied to all models to exclude the differences among the models originating from differences in meteorological and emission data. The meteorological data were used for initial, boundary, and nudging data or offline meteorological field. Furthermore, a horizontal grid with the same resolution as that of the meteorological data was adopted for all models. This setup enabled us to focus on model variability originating from the processes included in each model, for example, physical processes. The multimodel ensemble captured 40% of the atmospheric 137Cs events observed by measurements, and the figure of merit in space for the total deposition of 137Cs exceeded 80. The lower score of the atmospheric 137Cs than that of the deposition originated from the difference in timing between observed and simulated atmospheric 137Cs. Our analyses indicated that meteorological data were most critical for reproducing the atmospheric 137Cs events. The results further revealed that differences in 137Cs concentrations among the models originated from deposition and diffusion processes when the meteorological field was simulated reasonably well. The models with small deposition fluxes produced higher scores for atmospheric 137Cs, and those with strong diffusion succeeded in capturing the high 137Cs concentrations observed; however, they also tended to overestimate the concentrations. Key Points A model intercomparison of the atmospheric dispersion of 137Cs using identical input data was conducted Deposition and diffusion were key processes responsible for the differences among the models when the meteorological field was reproduced A score‐weighted ensemble‐mean distribution of the atmospheric 137Cs concentration was created