A real time forecasting model for landslides triggered by rainfall

The FLaIR model has been developed for the simulation and the forecasting of landslide movements activated by rainfall. It is composed of two modules. The first, rainfall-landslide module, correlates precipitation and landslide occurrence. The second, stochastic rainfall module, provides synthetic generation of rainfalls, giving a probabilistic representation of future precipitations. A mobility function, schematised as convolution of the rainfall intensity and a filter function, is related to the probability of landslide occurrence. The forecasting consists of the estimation at time tau of the value that the mobility function may attain to, at time t. Such a value depends on both the observed rainfall intensity, measured before tau, and the estimated one, derived from the stochastic rainfall module in the interval]tau, t]. Then the mobility function is composed of a deterministic and a stochastic part. In the paper a parameter, variance index, is introduced in order to describe the roles of the two components. For two very general classes of filters the analytical form of the variance index is determined providing an easy evaluation of the weights of the two components. The behaviour of different types of landslides is finally emphasised by two case studies.