REAL-TIME ESTIMATION OF HAZARD FOR LANDSLIDES TRIGGERED BY RAINFALL

Landslide movements triggered by rainfall can be foreseen in real-time by modelling the relationship between rainfall amount and landslide o ccurrence. This paper deals with the problem of the reliability of the FLaIR (Forecasting of Landslides Induced by Rainfalls) model when app lied to forecasting landslide movements in the usual condition of poor historical information availability. In this case, the identification of the admissibility field for the model parameters, instead of a poi nt estimation, leads to an improvement of the forecasting reliability. Moreover, this approach makes the model capable of taking into accoun t information embodied in periods of heavy rain but without movement. The concepts of informative content and foreseeability of landslide mo vements are introduced and their duality is analyzed. The effectivenes s of the estimation procedure described has been tested by application on two landslides located in southern Italy.