Algorithms for the Near-Real Time Identification and Classification of Landslide Events Detected by Automatic Monitoring Tools

Abstract

Early Warning Systems (EWS) represent one of the most effective activities for risk management and mitigation associated to landslides occurrence, allowing to reduce the possibility of human losses and damages to involved structures. One of the most challenging aspects of early warning activities is the reliable detection of potentially critical events starting from monitoring outcomes. In fact, a timely assessment of the landslide evolution would allow to undertake the most appropriate mitigation measures for a correct management of the ongoing event. In this context, it is especially important to avoid the occurrence of false alarms, which could induce significant issues from a social and economic point of view. The algorithm described in this study was conceived with the intent to identify an increasing pattern in landslide displacement rates and provide a classification of the detected event. The analysis is fully automated and relies on the elaboration of monitoring datasets sampled by automatic instrumentation to define the onset of acceleration of the event and assess its alert level. The proposed procedure was integrated in an automatic software developed for the elaboration of monitoring data sampled by automatic inclinometers, analyzing several datasets coming from different sites of interest with the objective of detecting the occurrence of unusual behaviors in the monitored landslides evolution.