On the Dynamic Response of Unstable Rock Slopes: Characterization, monitoring, and modeling based on ambient vibration data

Reliable mapping, characterization and monitoring of unstable rock slopes is a prerequisite for mitigation of risks associated with landslides to protect lives, settlements, and infrastructure. Most state-of-the art surveying techniques rely on sensing the Earth’s surface only, as drillings are expensive and spatially sparse measurements. Non-invasive geophysical techniques offer a wide range of possibilities to retrieve geological information at depth. In addition, they provide means to assess the hazard of coseismic slope failure, which is the most devastating secondary seismic hazard after tsunamis. In this thesis, I focus on recordings of ambient vibrations, which arise from distributed natural and artificial sources of vibration. The resulting continuous seismic signal can be measured at any time and location by seismometers. Measuring ambient vibration data allows for analyzing structural dynamic parameters of the unstable rock slope. These include the resonant frequency, polarization of vibration, energy dissipation (damping), seismic amplification, and seismic shear wave velocity. These parameters are proxies for properties of the slope instability and can be used to characterize and monitor a potential landslide. Rock stiffness and fracture properties govern the resonant frequency, while the orientation of fractures determines the polarization direction of the seismic wavefield. The amplification of seismic waves can reach factors exceeding ten within unstable slopes and strongly depends on the local fracture network. Therefore, amplification can be used to map the extent of the instability and simultaneously provide basic input parameters to model earthquake-induced slope failure. To confirm and analyze the normal mode behavior previously proposed for rock slope instabilities with deep compliant fractures, I apply different normal mode techniques to a set of unstable rock slopes by using field recordings and synthetic data. A key method is the Enhanced Frequency Domain Decomposition (EFDD) technique, which found broad application in civil engineering in the past decades. Besides the resonant frequencies and mode shapes, the technique also offers the possibility to retrieve the modal damping ratio, which is a measure for energy dissipation within and out of the system. In this work, I give first estimates of modal damping ratios in unstable rock slopes and provide potential explanations for the mechanisms driving this parameter. EFDD also enables the detection of close and higher order modes and can be applied as an efficient tool to map and structurally subdivide rock slopes instabilities. I implement the EFDD and other processing techniques for continuous real-time monitoring of unstable rock slopes to track the resonant properties and demonstrate how the seismic response of unstable slopes varies with changing environmental conditions and landslide kinematics. I monitor the dynamic response of three unstable rock slopes in Switzerland: the \textit{Frana del Valegion} instability near Preonzo, the Gemsstock northeast ridge in permafrost regime and the large and highly active Brienz/Brinzauls landslide. At the Brienz/Brinzauls landslide, the seismic amplification was observed to be elevated after periods of strong precipitation. This has direct implication for hazard assessment of coseismic slope failure, as the slope is more likely to be triggered by an earthquake during periods of increased amplification. The techniques and methodological approaches presented in this thesis have the potential for an efficient toolkit to characterize and monitor unstable slopes independent of surface displacement measurements. Therefore, they provide complementary techniques to conventional displacement measurements for landslide hazard assessment. The rapid and cost-effective mapping and subdivision of unstable rock masses could be especially beneficial in the early state of investigation to optimize subsequent monitoring systems and to plan more expensive in situ investigations