Journal of Geophysical Research: Earth Surface
© American Geophysical Union
Impact Factor: 3.426
ISI Journal Citation Reports © Ranking: 2014: 19/175 (Geosciences Multidisciplinary)
Online ISSN: 2169-9011
Associated Title(s): Journal of Geophysical Research
A probabilistic method of predicting landslides
Heavy rainfall events can lead to devastating landslides, but predicting when rainfall will cause a landslide is challenging. Most current landslide prediction methods consider past rainfall events that resulted in landslides, and then use that as input to provide a deterministic rainfall threshold for a landslide to occur. These methods simply predict either a landslide or no landslide for given rainfall conditions. In a new study, Berti et al. (2012) developed a probabilistic approach that returns a probability, from 0 to 1, of a landslide occurring for a given rainfall amount. They tested their approach on historical rainfall and landslide data from the Emilia-Romagna Region of Italy. They found that landslide occurrence was strongly related to duration, intensity, and total rainfall in a given event. The distribution of landslide probability showed a sharp increase at certain duration and intensity values, indicating a threshold where landslide becomes significantly more likely. However, antecedent rainfall, which had been thought to be an important factor in triggering landslides, was actually less important than the current rainfall event characteristics. The new probabilistic method could help significantly improve landslide forecasting.