REAL-TIME ESTIMATION OF HAZARD OF LANDSLIDE TRIGGERED BY RAINFALL

G. Iiritano, P. Versace (Dipartimento di Difesa del Suolo. Universita' della Calabria, 87040 Montalto Uffugo Scalo, Cosenza, Italy) &
B. Sirangelo (Istituto di Ingegneria Civile. Universita' di Salerno, 84084 Fisciano, Salerno, Italy

An automatic procedure to estimate the probability of exceeding a critical rainfall threshold for the initiation of landslides was implemented in the hydrological database METEOR. The procedure is based on the FLaIR model, an hydrological model to evaluate the hazard that a landslide-initiation rainfall threshold is exceeded. The model is based on the hypothesis that landslides occur when a mobility function, depending on the antecedent rainfall, exceeds a critical value. FLaIR forecasts the mobility function at time t* dividing the mobility function in a deterministic component, depending on the rain fallen before t (the starting time), and a stochastic component, depending on the rain that will fall in the intervall [t-t*]. The deterministic component is evaluated using the data recorded by a real-time network of rain-gauges. Various analytical formulations of the mobility function are implemented. The stochastic component is based on the forecast of the amount of rain that will fall in the future; it can be evaluated using different stochastic point models. For different forecast lag-times the software evaluates the minimum value of the stochastic component that exceeds the threshold of the mobility function. At the minimum value a probability of exceeding the threshold is given, based on the selected stochastic model. For simple models the probability is evaluated using an analytical expression. For more complex models the probability is estimated generating a synthetic series using the Montecarlo method. The system allows the real time evaluation of the landslide hazard at different forecast lag-times and could be employed in automatic landslide warning systems for civil protection purposes.