Early warning of rainfall-induced landslides based on empirical mobility function predictor

Abstract A stochastic real time predictor of rainfall-induced landslides has been developed. It couples the empirical slope stability model FLaIR with a point rainfall stochastic model. FLaIR model introduces a slope mobility function which links the occurrence of a slide movement to the characteristics of antecedent rainfall. Point rainfall external intermittence, namely the alternation of storms and dry periods, is modeled as an alternating renewal process (ARP). The properties of the ARP allow to assume that, during a storm, the future evolution of the mobility function depends only, in stochastic sense, on the hyetograph observed after the beginning of the ongoing storm (internal intermittence). Thus, the expected value of the mobility function is empirically evaluated by selecting, from the historical data set, only the storms with characteristics similar to the ongoing one. The predictor has been calibrated and validated on the basis of a nearly 48 years long hourly rainfall data record, collected by the rain gauge of Lanzo, in Northern Italy, close to the slope of Pessinetto, where six earth flows occurred during the observation period. The obtained results show that the proposed model provides reliable real time predictions of the slope mobility function up to a lead time of 6 h. The proposed predictor has been also tested as a part of an early warning system against earth flows to be operated at the slope of Pessinetto, by defining two threshold values of the mobility function, corresponding to alert and alarm levels, respectively. The obtained results show that, by properly setting the levels of probability of exceeding the two thresholds, at which the corresponding messages are launched by the system, it is possible, with a low number of false and missing messages, to gain some hours for effectively activating risk mitigation procedures.