LANDSLIDE HAZARD PREDICTION USING A 200-MEMBER ENSEMBLE FOR THE JULY 2020 HEAVY RAINFALL IN KYUSHU

Abstract

 While ensemble weather forecasting has been increasingly used in flood forecasting, there are few studies of applied to landslide prediction. In this study, the soil water index was calculated using 200 members rainfall forecasts to predict the risk of landslides for the line-shaped rainband that hit the Kyushu region in July 2020 heavy rainfall. The spatial distribution of the soil water index was visualized, and variations in values and locations were classified and confirmed. The soil water index was calculated at five different points of 1 km² in Ashikita City, where many landslides have occurred. As a result, 19% of the 200 members exceeded the alert level 4, which is equivalent to the evacuation order, a small value compared to the flood forecast where 56% of the members exceeded the previous maximum value. For the prediction of landslides, which are localized phenomena, the results suggest that landslide forecasting does not overlook low probability but high-risk values and takes into account the uncertainty in the location of the rainfall forecast.