A Prelimilary Study of the Rainfall Threshold and Early Warning System for Landslide in Taiwan

Catastrophic landslides and debris slides triggered by intense storms such as Typhoon Morakot (2009) occurred more frequently in recent years, and caused many casualties and economic loss in Taiwan. For the purpose of reducing damage resulted from geological hazards, this study collected landslide inventory which contains the information of occurrence such as time, location, size of failure, mean rainfall intensity, and rainfall duration with new measurement techniques. The precipitation data, satellite image, geology, terrain unit, topography mapping, and field survey are then combined to GIS for extracting the triggering factors of landslide. Furthermore, the multi period event-based landslide inventory and corresponding data are performed to establish the shallow landslide prediction model by logistic regression method and critical rainfall threshold for landslide early-warning system also developed. The landslide susceptibility model, landslide hazard scenario technique, and early-warning system were developed based on the systematical event-based landslide inventory. Validation with more than 200 cases show that the accuracy of the model is about 75-85% and the result could be applied to regional land-use planning, construction site selection, disaster prevention, and mitigation for civil infrastructure.