Chihiro KATO

A Study on Time-Series Pattern Analysis of Widespread Heavy Rainfall and Disaster Occurrence Risks

Toshiro KUMAKURA

This study aims to reevaluate the conventional rainfall classification method and analyze the characteristics of time-varying rainfall using radar 10-minute interval rainfall intensity data. Due to climate change, the frequency and intensity of heavy rainfall have increased, raising the risk of landslides and river flooding. Ground-based observation methods, such as AMeDAS, are limited in capturing localized rainfall changes, so radar data was used. The study focused on heavy rainfall events from August 2022, classifying rainfall patterns based on 10-minute interval data. The existing classification failed to reflect the characteristics of short-duration heavy rainfall, classifying many events as the highest hazard level (No. 5). By redefining classification criteria, more accurate rainfall patterns were obtained.
This study examined the relationship between rainfall pattern classification and landslide occurrence and found a link between local rainfall intensity and the location of landslides. The study also suggested that rainfall patterns in the upper reaches of rivers may influence disasters in the downstream areas. This highlights the need to incorporate geomorphic factors into disaster risk assessment. The findings of this study may contribute to the improvement of rainfall pattern classification using radar data and improve the accuracy of disaster prevention information in mountainous areas and areas where AMeDAS has not yet been introduced. Future work should focus on further analysis of rainfall patterns taking into account the topographical characteristics of river basins. Another future research topic is to explore the possibility of combining radar data with other types of environmental data in order to improve forecasting models for disaster risk assessment.