Predicting Rainfall-Induced Shallow Landslides in a Basin with a Water Content Index

Abstract

台灣及日本現行的土砂災害警戒系統均採用以雨量為指標的統計模式建置，惟此模式精度較差，且無法提供明確資訊，因此未獲地方政府及民眾信任。本研究採用斜坡單元為崩塌預測單元，整合物理模式法及多元迴歸分析，並以土壤水含量為指標，提出一新型態的集水區崩塌預測模式。此模式不僅可預測崩塌地點、時間及規模，亦有高速運算能力，同時也能估算坡面之降雨-逕流量，未來如整合其他模式，將可作為發展複合型土砂災害警戒系統之基礎。

The existing sediment disaster warning systems in Taiwan and Japan are established using a rainfall index based on a statistical model. However, because of the low warning hit rate and lack of definite warning information, they are not fully trusted. This study proposes a new landslide prediction model which adopts slope units as stability analysis targets and integrates a physically-based model and multiple regression analysis as well as used water content as an index for landslide prediction in a basin. The new model not only predicts the location, occurrence time, and scale of landslides but also offers high performance calculation. Moreover, it can also estimate the rainfall-runoff on a slope. If other sediment-related models are incorporated in the new model, it can serve as the foundation for developing a multi-hazard warning system.