Cluster analysis for dispersion patterns near nuclear power plants in South Korea

  Clustering analysis using air parcel trajectories is actively used to investigate transport patterns of pollutants. To estimate the impact of nuclide dispersion from nuclear accident, comprehensive information based on long-term meteorological data is required to eatablish a complete and efficient public protection plan. Most of nuclear plants in South Korea are located in a complex terrain near coastal area that involves complicated meteorological phenomenon such as sea breezes and mountain-valley breezes. Robust approach based on long-term climatrological data is required to fully resolve the impacts near Korean nuclear power plants.

  In this study, we assessed the impacts of potential nuclear accident in South Korea by clustering dispersion patterns using 10-year meteorological data. Flow patterns are clustered using trajectory cluster analysis, and then combined with dispersion simulations to demonstrate the clustered dispersion patterns by each season and nuclear power plant.

  The long-term meteorological simulations from 2007 to 2016 were used to evaluate the potential impact of nuclear accidents in Korea, and the modeling framework was designed to show the impact map according to the flow patterns near each nuclear power plant. NOAA HYSPLIT modeling additional clustering analysis suggests that two or three cluster patterns for each power plant can be used. A total of 38 flow patterns are classified near the four nuclear plants in the previous season based on a 10-year wind field analysis. Korea has very complex terrain and coastal areas, and more sophisticated modeling efforts are needed to fully understand the more realistic dispersion characteristics of air masses. In terms of space-time resolution, updating land use information for simulation is very important for weather simulation near the surface of Korea.

  The results of this study can be used as a guideline for constructing a modeling framework for nuclide diffusion simulations, but given these complex simulation configurations, the results demonstrated in the current study are should be interpreted with caution.