he Rainfall Triggering Index (RTI) model is adopted to set up the critical rainfall for each township since 2005 in Taiwan, and the debris-flow warning system based on the RTI model is successful in reducing casualties.
However, the antecedent rainfall calculation using the deduction coefficient of "t" days in the RTI model leads to a false alert rate higher under some rainfall patterns (e.g., long-term duration and lower rainfall intensity). This study suggests a modified method to calculate the antecedent rainfall and effective accumulated rainfall to solve the above-mentioned problems. We also establish a new warning model, which uses the QPESUM data for the past decade, the identified results of remote-sensing image, and the disaster records, to predict landslides. In the case study of the Wulai District during Typhon Soudelor in 2015, the new warning model offers good prediction of the times and locations of landslides. This study also proposes a new platform which displays the rainfall-induced disaster hot zones. These findings
can help government officials to make appropriate decisions during emergency response.
Key Words: Landslide, debris flow, warning system, QPESUMS, effective accumulated rainfall.