Landslide time-of-failure forecast and alert threshold assessment: A generalized criterion

Abstract The forecast of a landslide's time of failure and the definition of alert thresholds are fundamental aspects in the study of natural hazards. However, these tasks are particularly difficult due to the large number of parameters and factors involved, and are therefore usually performed with a site-specific approach. This work describes an attempt to generalize the behavior of a landslide approaching collapse, with particular attention to the definition of a general criterion to define alert thresholds. The procedure started with the creation of a database of displacement data recorded for historical landslides, then the inverse velocity model was applied to these datasets to evaluate the time of failure under the assumption of linear behavior during the accelerating phase. A model calibration was conducted to best describe the monitored data and highlight any non-linear trend. A curve describing the velocity versus time relationship was then computed for each single slope failure case using the parameter obtained through this operation. In the final step of the study, these curves were processed with a normalization procedure, thus obtaining a dimensionless velocity-related coefficient. This parameter allowed the comparison of different landslide datasets on a single graph, which can be used as a general reference to define alert thresholds for emergency purposes. In order to test the criterion's ability to represent landslide behavior, the procedure was also applied to a different case by simulating progressive data acquisition.