Modelling the relationships between rainfall, hydrology and deep-seated landslides on a real-time basis and with sufficient response time constitutes a key challenge of early warning and decision support systems. We show here our findings on an autonomous, real-time modelling of slope kinematics at Veslemannen slope (SW Norway), aiming to determine slope-response to precipitation by considering time-dependent aspects of slope behavior. Answering this question will not only increase our understanding of the slope, but also help on the hazard management prior to a potential collapse.
Rock Slope Failures (RSF) around the world generally display creep strain curves characterized by episodic acceleration and deceleration phases in a so-called stick-slip pattern (Prior & Stephens 1972). A sudden increase in pore water pressure and increasing stresses in the landslide body is normally observed after each rainfall episode, revealing a fast infiltration through preferential paths (Corominas et al. 2005; Loew et al. 2012; Blikra et al. 2013; Crosta et al. 2014). This behaviour is normally followed by a progressive decay in velocity reflecting a gradual reduction of water pressure, possibly due to flow diffusion (Iverson 2000). When the movement enters into tertiary creep phase, strain rate is normally characterized by a continuous acceleration until sudden catastrophic failure (Petley et al. 2005; Rose & Hungr 2007; Agliardi et al. 2010). Degrading permafrost also controls the timing and location of RSF (Krautblatter et al. 2013).
Model parametrization in RSF forecasting is normally intricate due to inherently complex geological, hydrological and geomechanical conditions. In addition, epistemic uncertainties also appear due to progressive and/or abrupt changes of slope behaviour in a time-dependent manner (Eberhardt et al. 2004; Petley et al. 2005; Abellan et al. 2015). Calculating slope acceleration and deceleration phases as response to rainfall using inverse modeling and empirically-derived response functions has proven useful before (Belle et al. 2014; Abellan et al. 2015; Bernardie et al. 2015).