Abstract
Traditionally, the deterministic rainfall threshold has been widely used for early warning of landslides, which is often developed based on visual inspection of historical rainfall events in a region that triggered landslides. As there is no deterministic relationship between rainfall and the occurrence of landslides, it is not uncommon to observe in practice that a rainfall above the deterministic rainfall threshold may not cause landslides, and vice versa. In this paper, a numerical study is first conducted to gain insights on factors affecting the rainfall threshold for early warning of landslides. Then, a procedure based on the support vector machine is suggested to develop probabilistic rainfall thresholds, where the reliability associated with different rainfall thresholds can be explicitly quantified. Finally, two application examples are introduced to illustrate and validate the suggested method for developing rainfall thresholds probabilistically. The probabilistic rainfall thresholds presented in this study can provide more comprehensive information for issuing warning messages and risk management of landslide hazards.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
