Spatiotemporal slope stability analytics for failure estimation (SSSAFE): linking radar data to the fundamental dynamics of granular failure

Abstract

Impending catastrophic failure of granular earth slopes manifests distinct kinematic patterns in space and time. While risk assessments of slope failure hazards have routinely relied on the monitoring of ground motion, such precursory failure patterns remain poorly understood. A key challenge is the multiplicity of spatiotemporal scales and dynamical regimes. In particular, there exist a precursory failure regime where two mesoscale mechanisms coevolve, namely, the preferred transmission paths for force and damage. Despite extensive studies, a formulation which can address their coevolution not just in laboratory tests but also in large, uncontrolled field environments has proved elusive. Here we address this problem by developing a slope stability analytics framework which uses network flow theory and mesoscience to model this coevolution and predict emergent kinematic clusters solely from surface ground motion data. We test this framework on four data sets: one at the laboratory scale using individual grain displacement data; three at the field scale using line-of-sight displacement of a slope surface, from ground-based radar in two mines and from space-borne radar for the 2017 Xinmo landslide. The dynamics of the kinematic clusters deliver an early prediction of the geometry, location and time of failure.

Pinpointing Early Signs of Impending Slope Failures From Space

“…This pattern is transient initially but then ultimately persists in space and time close to and during failure. As reported in recent studies (Singh & Tordesillas, 2020; Tordesillas et al., 2021), such spatiotemporal clustering pattern has been discovered in a wide range of failures, despite their scales (laboratory versus slope versus field) and types (brittle versus ductile).…”

“…This pattern is transient initially but then ultimately persists in space and time close to and during failure. As reported in recent studies (Singh & Tordesillas, 2020;Tordesillas et al, 2021), such spatiotemporal clustering pattern has been discovered in a wide range of failures, despite their scales (laboratory versus slope versus field) and types (brittle versus ductile). The spatiotemporal dynamics of kinematic clustering can be directly related to the fundamental physics of granular media deformation and failure, especially to the process of force transmission (Tordesillas et al, 2021;.…”

“…That is, compared to links elsewhere, the links in the bottleneck transmit forces close to their capacity which is the force that needs to be overcome to break the link. In ; Tordesillas et al (2021), it was shown that the path that percolates through the body containing such vulnerable links, i.e., force bottleneck, is in fact the path of least resistance to failure and is where the macrocrack in small samples eventually developed at failure. In monitoring data, this path delimits the landslide boundary at the time of collapse (Tordesillas et al, 2021).…”

“…Application of these in the analysis of ground motion radar data gave promising results (Das & Tordesillas, 2019; Singh & Tordesillas, 2020; Tordesillas et al., 2018, 2021; Wang et al., 2020; Zhou et al., 2020). In particular, analysis of thousands of displacement time series data for a developing rockslide in an operational open pit mine, obtained using GB‐InSAR, gave a precise prediction of the ultimate location and geometry of the failure region, weeks in advance of the event (Tordesillas et al., 2018, 2021). Building on this capability, a subsequent stochastic spatiotemporal forecasting algorithm identified imminent failure 10 hr prior to the collapse (Wang et al., 2020).…”

Engineering-Geological Analysis of a Subaerial Landslide in Taan Fiord, Alaska

“…In the present study, the multi-temporal aerial radar data were used to reconstruct the slope morphology over time and to constrain the subsequent slope stability analyses via the limit equilibrium method. The applicability of remote-sensing data for landslide investigation has been confirmed in various studies [37][38][39][40][41][42][43]. The aim of this study is to gain a better understanding of the landslide process, and of which factors or external triggers may have induced the slope failure.…”