A new method for characterizing and forecasting the evolution of slope moviments

Abstract

Characterizing and forecasting the evolution of slope movements are highly studied topics due to the possible consequences of landslides on buildings, infrastructures and in terms of loss of life. The international scientific literature provides detailed analysis and modelling of specific phenomena, whereas a generalized procedure aimed at identifying common properties of landslides is lacking. In this thesis a general framework for understanding typical behaviours of slope movements is presented, with the aim of classifying stages of activity of landslides and predicting their time evolution. For this purpose, monitoring displacement data are collected for several case studies from literature that differ in materials, geometry and triggering factors. These data are analysed referring to single activity stages, selected on the basis of geometric properties of data series and on variations of triggering causes. Thereafter, displacement dimensionless diagrams are constructed, allowing the identification of some trends, each characterized by common growth properties related only to the evolutive stage not depending on the specific characteristics in terms of dimensions, geometry, materials and so on. Dimensionless trends are analysed through the derivatives of displacement function up to the third order, namely velocity, acceleration and jerk. Sign and growth properties of derivatives allow identifying common characteristics in the evolution of landslides. Hence, an appropriate interpolating function of displacement data is introduced in order to develop a forecasting algorithm able to establish alert levels useful in early warning systems. This procedure is applied to the cases from literature and to La Saxe rockslide (Courmayeur, Valle d’Aosta, Italy) for which continuous monitoring data are available with a sampling frequency of one hour. Dimensionless trends are obtained in terms of displacements, velocities, accelerations and jerks, showing a satisfying correlation with actual time evolution of landslide motions. In addition the forecasting algorithm is applied to an important failure that locally involved La Saxe rockslide and to some instabilities in an open pit mine, showing the capability of the method in establishing useful alert levels. [edited by Author]