Massimo Melillo, Maria Teresa Brunetti, Stefano Luigi Gariano, Silvia Peruccacci, Fausto Guzzetti, 2015, AUTOMATIC RECONSTRUCTION OF RAINFALL EVENTS AND OBJECTIVE DEFINITION OF RAINFALL THRESHOLDS FOR LANDSLIDE OCCURRENCE, Le Giornate dell'Idrologia, Perugia, 06/10/2015,

Objective criteria for the identification of rainfall events are ambiguous or subjective, particularly when dealing with the forecast of rainfall-induced landslides. In an attempt to overcome the problem, we developed an algorithm for the objective and reproducible reconstruction of rainfall events, and of rainfall conditions responsible for landslide occurrence (Melillo et al., 2015). The algorithm is implemented in a code written in the R language, and comprises three distinct modules to perform_ (i) the reconstruction of distinct rainfall events, in terms of rainfall duration (D in h) and cumulated event rainfall (E in mm), (ii) the identification of one or more rainfall ED conditions responsible for observed landslides, and (iii) the definition of rainfall thresholds for possible landslide occurrences. The algorithm elaborates rainfall records and reconstructs individual rainfall events using pre-defined parameters to account for different geographical and climatic conditions. A comparison between rainfall characteristics in different regions can be performed. We applied the algorithm in Sicily, southern Italy, using (i) rainfall measurements obtained from a network of 169 rain gauges, and (ii) information on 265 rainfall induced landslides occurred between January 2002 and December 2012. The algorithm reconstructed 29,270 rainfall events, corresponding to a mean of 24 rainfall events per year. For each rain gauge, the algorithm calculated statistics for the reconstructed rainfall events. The statistics have been proved useful for geographical and seasonal analysis of rainfall patterns. Next, using information on landslide occurrence, the algorithm reconstructed 472 rainfall ED conditions as possible triggers of the observed failures. Then, the algorithm exploited the multiple rainfall conditions to define objective and reproducible empirical rainfall thresholds for the possible initiation of landslides. The calculated thresholds can be implemented in an operational early warning system for shallow landslide forecasting.

Data from