Radioactive Nuclei Emission Analysis from Fukushim Dai-ichi Nuclear Power Plant by Inverse Model

A lot of radioactive materials were emitted in the accident of the Fukushima Daiichi nuclear power plant, which occurred in March 2011. Although various organizations in and outside Japan released the prediction result using a transport model about this accident. The most case assumed a fixed emission value and resulted that the prediction accuracy was relatively limited. The important feature of this accident is that the source position is clearly known. This means that only time and vertical emission variations are not known (in this case, we also know that the emission height is not so high altitude). In such a case, the technique of inverse model was powerful tool to answer the questions. This technique are widely used in carbon cycle analysis and recently used in volcanic eruption and dust aerosol emission analysis. In such a case, it is expectable that we could obtain high resolution and more precise analysis by using prior emission information with relatively low computational costs. We used tagged simulation results by global aerosol model named MASINGAR (Tanaka et al., 2005). The horizontal resolution is TL319 (about 60 km). We released three tagged tracers (Cs137) from every three levels from the surface. We released them every three hours with 1Tg/hr (finally, we accumulated daily mean). We collected 50 sites' daily observation data in the world (CTBTO, Ro5, Berkeley and so on). The analysis period is 40 days from March 11 – April 19. The prior emission information are JAEA posterior (Chino et al., 2011) and NILU prior (not posterior) (Stohl et al., 2011) as our observation data are almost similar to latter study. We examined several sensitivity tests by changing observation data and prior emission uncertainties. We find that estimated total emission amount from Mar. 11 to Apr. 19 is about 18PBq. Our results indicated that JAEA analysis (9PBq) is too small and NILU analysis (28PBq) is too large. We also find that the maximum radio nuclei emission is occurred during 15 March. The precision of our analysis highly depends on some element (quality and quantity of observational data, precision of transport model). By developing this system, it is expectable for construction of the near-real-time forecasting system at the time of the occurrence of an accident of a nuclear power plant, volcanic eruption and so on. To achieve this system, we should reconstruct observation data network by OSSE (observation system simulation experiment). We also consider it possible to obtain robust result by using multi-model (processes) ensemble results with inverse model. The results of this study are available for modification of many processes of aerosol transport models.