Risk Analysis of Heavy Rainfall and Associated Disaster Occurrence Based on Pattern Recognition of Meteorological Fields Using Self-Organizing Map

Koji Nishiyama, Toshiyuki Moriyama, Katsuhiro Morishita, Shinobu Izumi, Izumi Yokota, Kenichi Tsukahara

Wednesday 1 july 2015

12:36 - 12:39h at South America (level 0)

Themes: (T) Flood risk management and adaptation, (ST) Flood risk assessment, Poster pitches

Parallel session: Poster pitches: 9J. FloodRisk - Assessment

In Japan, heavy rainfall events have been caused by the supply of a large amount of water vapor into the Japan Islands. Consequently, such heavy rainfall events have led to serious disasters due to intense debris flow or river flood. Therefore, it is very important to clearly identify what kind of meteorological situations have a potential to cause heavy rainfall. For solving such a problem, this study classified non-linear complicated long-term meteorological fields including western Japan into visually-detectable patterns on two-dimensional arrays using SOM algorithm, and investigated relationships between meteorological field patterns and heavy rainfall frequency. In the next analysis, this study investigated features of meteorological field patterns constructed by the SOM algorithm, focusing on actual disaster case that occurred in Izumi city of Kagoshima prefecture in Japan. From these analysis, it could be clearly and visually identified what kinds of meteorological field patterns have a potential to cause heavy rainfall. Moreover, it was found that heavy rainfall and associated debris flow disaster that occurred in Izumi city was caused by meteorological field patterns characterized by the formation of the high PW region with Low Level Jet. It was also found that these meteorological field patterns have high probability to cause heavy rainfall. Therefore, it could be highly expected that the SOM results are easily and visually available for the diagnosis of whether or not actual meteorological situations will have a potential to cause heavy rainfall if the input of the meteorological situations into the SOM system is conducted.