Approaches of data analysis from multi‐parametric monitoring systems for landslide risk management

In the last decades, several approaches were proposed accounting for early warning systems to manage in real time the risks due to fast slope failures where important elements, such as structures, infrastructures and cultural heritage are exposed. The challenge of these approaches is to forecast the slope evolution, thus providing alert levels suitable for managing infrastructures in order to mitigate the landslide risk and reduce the “response” time for interventions. Three different strategies can be defined in this regard: an Observation‐Based Approach (OBA), a Statistic‐Based Approach (SBA) and a Semi‐Empirical Approach (SEA). These approaches are focused on searching relations among destabilizing factors and induced strain effects on rock mass. At this aim, some experiments are being performed at different scales in the framework of consulting activities and research projects managed by the Research Centre for the Geological Risk (CERI) of the University of Rome “Sapienza”. These experiments are testing different kind of sensors including extensometers, strain gauges, rock‐thermometers, interferometers, optical cams connected to Artificial Intelligence (AI) systems, for detecting changes in rock properties and detecting stressstrain changes, as well as pluviometers, anemometers, hygrometers, air‐thermometers, micro‐ or nano‐ accelerometers and piezometers for detecting possible trigger of deformational events. The results of this Ph.D. thesis demonstrate that the data analysis methods allowed the identification of destabilizing actions responsible for strain effects on rock mass at different dimensional scale and over several time‐window, from short‐ to long‐ period time scale. Furthermore, the three approaches were to be suitable to recognize precursor signals of rock mass deformation and demonstrated the possibility to provide an early warning.