The Impact of Surface Wind Data Assimilation on the Predictability of Near-Surface Plume Advection in the Case of the Fukushima Nuclear Accident

Abstract

We investigated the predictability of plume advection in the lower troposphere and the impact of AMeDAS surface wind data assimilation by using radioactive cesium emitted by the Fukushima nuclear accident in March 2011 as an atmospheric tracer. We conducted two experiments of radioactive plume predictions over eastern Japan for March 15, 2011 with a 3-km horizontal resolution using the Japan Meteorological Agency non-hydrostatic weather forecast model and local ensemble transform Kalman filter (JMANHM-LETKF) data assimilation system. The assimilated meteorological data were obtained from the standard archives collected for the Japan Meteorological Agency operational numerical weather prediction and the AMeDAS surface wind observations. The standard archives do not contain land-surface wind observations. The modeled radioactive cesium concentrations were examined for plume arrival times at 40 observatories. The mean error of the plume arrival times for the standard experiment (assimilating only the standard archives) was 82.0 min with a 13-h lead-time on an average. In contrast, the mean error of the AMeDAS experiment (assimilating both the standard archives and AMeDAS surface wind observations) was 72.8 min, which was 9.2 min (11 %) better than that of the standard experiment. This result indicates that the plume prediction has a reasonable accuracy for the environmental emergency response and the prediction can be significantly improved by the surface wind data assimilation.

Deposition and Dispersion of Radio‐Cesium Released Due to the Fukushima Nuclear Accident: Sensitivity to Meteorological Models and Physical Modules

“…Since both the NHM and NHM‐LETKF also employ the same Mellor‐Yamada‐type PBL scheme (MYNN30) as the WRF simulations, the differences in the simulated surface air concentrations due to differences in vertical diffusivities were not significant. Within the PBL, all of the simulations were not assimilated to the meteorological analysis data, except for NHM‐LETKF: the surface wind data of the JMA's Automated Meteorological Data Acquisition System at more than 200 stations in the model domain were assimilated to the simulation (Sekiyama et al, ). This partly explains why NHM‐LETKF‐Chem performed best in the quantitative simulation of the surface activity concentrations of 137 Cs.…”

“…The CTM part of NHM-Chem solves for tropospheric photochemical reactions and aerosol dynamics processes, but a simplified version was built and applied for this study. A previous version of the CTM part was called the Regional Air Quality Model 2 (RAQM2; Kajino et al, 2012), and its simplified version (Adachi et al, 2013) was also applied to previous studies of the atmospheric simulations of the radionuclides released in association with the FDNPS accident (Adachi et al, 2013;Sekiyama et al, 2015Sekiyama et al, , 2017).…”

“…e This is a meteorological analysis. f LETKF: Local ensemble transform Kalman filter (Kunii, 2013;Sekiyama et al, 2017). g WRF: Weather Research and Forecast model (Skamarock et al, 2008).…”

“…As shown in Figure 2a, the simulation domain and grid resolution are the same as those used in our previous studies (Adachi et al, 2013;Sekiyama et al, 2015Sekiyama et al, , 2017. There are 213 × 257 horizontal grid cells with a 3km grid resolution on the Lambert conformal coordinate.…”

“…Within the PBL, all of the simulations were not assimilated to the meteorological analysis data, except for NHM-LETKF: the surface wind data of the JMA's Automated Meteorological Data Acquisition System at more than 200 stations in the model domain were assimilated to the simulation (Sekiyama et al, 2017). This partly explains why NHM-LETKF-Chem performed best in the quantitative simulation of the surface activity concentrations of 137 Cs.…”