Diagnosis of Synoptic Weather Patterns Causing Heavy Rainfall Occurrence Using Self-Organizing Maps

Sammanfattning: The annual number of events with precipitation over 50 mm/h have increased from 1970s and this trend is likely to be continued due to climate change in Kyushu, Japan. Therefore, it is important to recognize what kind of meteorological fields have contributed to the occurrence of heavy rainfall events and to reveal whether it is possible to diagnose heavy rainfall risk. Firstly, this study combines the Self-Organizing Map (SOM) and Radar/Rain gauge analyzed precipitation to provide the distribution of heavy rainfall frequency on the two-dimensional map. As a result, 19520 meteorological fields observed for 40 years are classified into 40 synoptic weather groups and the top 10 groups are characterized by four reasons, the existence of 1) strong southwest wind and large amounts of precipitable water (PW), 2) counterclockwise circulation with large PW, 3) tropical cyclone and 4) stationary front. Secondly, the Global Spectral Model is combined with the structured SOM for diagnosing the probability of heavy rainfall occurrence within a range of few days. Following the case studies, the probability of heavy rainfall occurrence can be increased and stabilized around 36 h before heavy rainfall events. Furthermore, the SOM can relate the diagnosed patterns to historical rainfall events.