Ensemble dispersion simulation of the radioactive aerosol emitted from the Fukushima Daiichi nuclear power plant

We conducted ensemble simulations for dispersion and deposition of the radioactive aerosol emitted from the Fukushima Daiichi nuclear power plant (FDNPP) in March 2011. The ensemble of meteorological variables were prepared by an ensemble-based data assimilation system, which consisted of the Japan Meteorological Agency non-hydrostatic weather-forecast model (JMA-NHM), the four-dimensional local ensemble transform Kalman filter (LETKF), and the JMA operationally-used observation datasets. The data assimilation was implemented every 3 hours (= assimilation time window) with a 3-km horizontal resolution using 20 ensemble members. Then, a regional radiochemistry transport model (RRTM) was driven by the meteorological variables of each ensemble member. Additionally, a deterministic RRTM simulation was conducted with the ensemble average of the meteorological variables. These model simulations were compared and validated with the time series of radioactive Cesium-137 concentrations measured by the Japan Atomic Energy Agency (JAEA) at Tokai Village approximately 100 km south of the FDNPP. Figure 1 shows the Cesium-137 concentrations at the model surface layer of the ensemble runs and the deterministic run from 21:00 March 20 to 21:00 March 21 local time (LT), in which the meteorological variables were interpolated by 3-hour model forecasts through the assimilation time window. Figure 2 shows the percentile of the 20 ensemble distributions, in which the surface Cesium-137 concentration was above the threshold (= 15 Bq/m3), with the contour line (= 15 Bq/m3) of the deterministic run. The cross mark in Fig.2 illustrates the location of Tokai Village. The local times of these snapshots are 4:00, 8:00, and 12:00 on March 21. Figure 1 indicates that the deterministic simulation has limited capacity for predicting the location and intensity of Cesium-137 plumes. In reality, the Cesium-137 concentration sharply rose at 4:00 LT and reached more than 400 Bq/m3. However, the deterministic simulation delayed the peak time by 2 hours and weakened the peak concentration by 100 Bq/m3. In contrast, some of the ensemble members successfully reproduced the reaching time and intensity of the Cesium-137 plume, although some of the other members performed worse than the deterministic simulation. In this case, Fig. 2 clearly shows a large ensemble spread of the Cs-137 distribution. The high percentile area of the Cesium-137 plume (above the threshold concentration) was very narrow while the low percentile area spread broadly upwind and downwind. Unfortunately, the model uncertainty of aerosol dispersion and deposition is often large even if the errors of meteorological analyses are relatively small. Nevertheless, the ensemble simulation can provide probabilistic information on radioactive aerosols, e.g., a small number of possible outcomes or severe scenarios. As far as we know, this study is the first time that a full-fledged ensemble simulation has been conducted with high (= mesoscale) model resolution using a chemistry transport model and well-prepared (= statistically plausible) meteorological ensemble members. Furthermore, the successful ensemble simulation in this study will develop into a simultaneous LETKF data assimilation of Fukushima radioactive aerosols and meteorological observation data.