Search

As increased pore pressure due to rainwater infiltration is the primary triggering factor for landslides, rainfall is normally established as the first condition that allows definition of the predictability of their occurrence over the short term for a determined location. It is therefore common to propose landslide occurrence predictions based on awareness of accumulated background rainfall from historic data analysis. This paper proposes that landslide occurrence be predicted by quantitatively estimating their probabilities based on accumulated rainfall during a pair of time frames (e.g., 24 and 72 h) prior to the event, with selection determined by statistical dependence readings. This method was applied to the Quitandinha river basin region in the municipality of Petrópolis, Brazil, using background data from 2003 to 2009. It was observed that landslide occurrence presented the highest relevance level with the two accumulated rainfall periods of 24 and 96 h, and it was possible to estimate the probability of occurrence of at least one, three, or five landslides depending on the accumulated rainfall rates during these time frames. [ABSTRACT FROM AUTHOR]