January, 2013
The NOAA Air Resources Laboratory’s Roland Draxler participated on a World Meteorological Organization (WMO) Task Team to evaluate the meteorological aspects of the Fukushima Daiichi nuclear power plant accident. The Task Team focused on how the meteorological analyses could be improved for use in assessment studies that utilize atmospheric transport, dispersion, and deposition models (ATDM). The Team recently completed its final report summarizing its findings. In addition to NOAA, other contributors to the report included the U.K. Met Office, the Japan Meteorological Agency, the Meteorological Service of Canada, and the Austrian Meteorological Service.
The results showed that some of the best deposition and air concentration results were obtained from models that used the highest spatial and temporal resolution meteorological fields. However, it was also found that the use of high resolution spatial and temporal observed precipitation data, in place of the model, resulted in a deterioration of the deposition and air concentration calculations. The most consistent results were obtained when some of the better performing models were combined into a multi-model ensemble.
Background: Various assessment studies are being conducted over the next several years to evaluate the consequences of the radionuclide releases. The WMO organized a team from member countries that had expertise with respect to meteorological data, as well as ATDM simulations. Each participant provided meteorological data and ATDM simulations for evaluation.
Significance: The accidental or intentional release of chemical, biological or nuclear agents can have significant health, safety, security, economic, and ecological implications. Following the accident at the Fukushima Daiichi nuclear power plant, not all dispersion- meteorology model combinations had the same performance at all locations and time periods. To address this issue, a multi-model ensemble approach was devised. This provides for the computation of uncertainties associated with the model predictions, as well as a more consistent level of performance for both deposition and air concentration. The results suggest that future assessment studies, as well as real-time incident response modelling, would be improved by the use of multi-model ensembles rather than relying upon a single deterministic prediction.
For More Information, contact: Roland Draxler