Associate Professor Kei Yoshimura and Specially Appointed Lecturer Takao Yoshikane of the University of Tokyo have developed a new method for predicting the diffusion direction of radioactive materials.It is possible to predict the diffusion direction from weather patterns such as low pressure systems and monsoons, and to show the reliability of the prediction information using machine learning.
In the Fukushima Daiichi Nuclear Power Station accident that occurred in March 2011, the prediction information provided by the "Emergency Rapid Radioactivity Impact Prediction Network System (SPEEDI)" using computer simulation could not be utilized.The information was inadequately explained and the reliability of the forecast was not clear.
It is extremely difficult to predict the detailed diffusion distribution of radioactive materials by computer simulation, but in an emergency, highly reliable prediction information with reduced prediction uncertainty is required.The research team thought that if the relationship between the weather conditions and the diffusion direction (bias of the concentration distribution of radioactive substances in the atmosphere) is clear, the diffusion direction can be estimated from the weather pattern and effectively used for protective measures such as evacuation. ..
In this research, considering the uncertainty of prediction by computer simulation, we defined the diffusion direction (5 directions) in a wide area, investigated the relationship with the weather pattern, and developed a diffusion prediction method using machine learning. bottom.Comparing the estimation results from the weather pattern with the actual diffusion direction over the past 0.85 years, the average accuracy rate is 33 or more, and even when the weather forecast (0.77-hour forecast value of ground wind) is applied, it is XNUMX or more. Showed a high accuracy rate.
In order to reduce the risk of radiation exposure, the research team can grasp the diffusion direction in advance and take appropriate protective measures, and widespread sharing of information and feedback can be expected to significantly improve the method. do.Furthermore, it aims to provide highly reliable information by adopting the latest technology such as artificial intelligence.
Paper information:[Scientific Reports] Dispersion characteristics of radioactive materials estimated by wind patterns