A Storm-Triggered Landslide Monitoring and Prediction System: Formulation and Case Study

Abstract

Abstract Predicting the location and timing of mudslides with adequate lead time is a scientifically challenging problem that is critical for mitigating landslide impacts. Here, a new dynamic modeling system is described for monitoring and predicting storm-triggered landslides and their ecosystem implications. The model ingests both conventional and remotely sensed topographic and geologic data, whereas outputs include diagnostics required for the assessment of the physical and societal impacts of landslides. The system first was evaluated successfully in a series of experiments under idealized conditions. In the main study, under real conditions, the system was assessed over a mountainous region of China, the Yangjiashan Creeping (YC) slope. For this data-rich section of the Changjiang River, the model estimated creeping rates that had RMS errors of ∼0.5 mm yr−1 when compared with a dataset generated from borehole measurements. A prediction of the creeping curve for 2010 was made that suggested significant slope movement will occur in the next 5 years, without any change in the current precipitation morphology. However, sliding will become imminent if a storm occurs in that 5-yr period that produces over 150 mm of precipitation. A sensitivity experiment shows that the identified location fails first, triggering domino-effect slides that progress upslope. This system for predicting storm-triggered landslides is intended to improve upon present warning lead times to minimize the impacts of shallow, fast moving, and therefore hazardous landslides.