Search

We study the physical mechanisms that drive alpine slope deformation during water infiltration and depletion into fractured bedrocks. We develop a fully coupled hydromechanical model at the valley scale with multiscale fracture systems ranging from meter to kilometer scales represented. The model parameterized with realistic rock mass properties captures the effects of fractures via an upscaling framework with equivalent hydraulic and mechanical properties assigned to local rock mass blocks. The important heterogeneous and anisotropic characteristics of bedrocks due to depth-dependent variations of fracture density and stress state are taken into account and found to play a critical role in groundwater recharge and valley-scale deformation. Our simulation results show that pore pressure actively diffuses downward from the groundwater table during a recharge event, rendering a critical hydraulic response zone controlling surface deformation patterns. During the recession, the hydraulic front migrates downwards and the deformation recorded at the surface (up to ~4 cm) rotates accordingly. The most essential parameters in our model are the fracture network geometry, initial fracture aperture (controlling the rock mass permeability), and regional stress conditions. The magnitude and orientation of our model’s transient annual slope surface deformation are consistent with field observations at our study site in the Aletsch valley. Our research findings have important implications for understanding groundwater flow and slope deformations in alpine mountain environments., ESS Open Archive

Solid earth : SE 13(2), 301-322 (2022). doi:10.5194/se-13-301-2022, Published by Copernicus Publ., Go��ttingen

In deglaciating environments, rock mass weakening and potential formation of rock slope instabilities is driven by long-term and seasonal changes in thermal- and hydraulic boundary conditions, combined with unloading due to ice melting. However, in-situ observations are rare. In this study, we present new monitoring data from three highly instrumented boreholes, and numerical simulations to investigate rock slope temperature evolution and micrometer-scale deformation during deglaciation. Our results show that the subsurface temperatures are adjusting to a new, warmer surface temperature following ice retreat. Heat conduction is identified as the dominant heat transfer process at sites with intact rock. Observed nonconductive processes are related to groundwater exchange with cold subglacial water, snowmelt infiltration, or creek water infiltration. Our strain data shows that annual surface temperature cycles cause thermoelastic deformation that dominate the strain signals in the shallow thermally active layer at our stable rock slope locations. At deeper sensors, reversible strain signals correlating with pore pressure fluctuations dominate. Irreversible deformation, which we relate with progressive rock mass damage, occurs as short-term (hours to weeks) strain events and as slower, continuous strain trends. The majority of the short-term irreversible strain events coincides with precipitation events or pore pressure changes. Longer-term trends in the strain time series and a minority of short-term strain events cannot directly be related to any of the investigated drivers, but changes of glacial loading and fatigue processes are considered as potential causes., Earth and Space Science Open Archive

The Flims rockslide/rock avalanche (FRRA) is the largest long runout landslide in Europe. This event provides a unique opportunity to study the pre-failure and failure behavior of a large rock slope, as both the source zone and deposit of this event are accessible. In this study, we perform engineering geological and geomorphic field mapping as well as stability and runout modeling in order to explore the preconditioning and triggering factors that resulted in failure of this event, and to infer the mechanisms that governed its runout. By combining these analyses, we qualitatively comment on the mechanisms that lead to the transition from a rockslide to a long runout, catastrophic rock avalanche. Our engineering geological and geomorphic field mapping has revealed that the FRRA failed along a sliding zone that features numerous, large scale steps. Previous work at the site, as well as new analysis of thin sections, has revealed the presence of marl-like layers within the failed stratigraphic unit. Our stability analysis shows that the presence of low strength layers at the depth of the rupture surface is required for failure to initiate, and that failure could be triggered either by strong seismic shaking, elevated pore-water pressures, or a combination of both. The results of the runout analysis show that this event likely remained coherent for a large portion of its motion, and that liquefaction of alluvial sediments at the toe of the slope may have enhanced the runout distance of this rock avalanche. Combining the mapping, stability and runout modeling has shown that the basal shear strength required for the runout analysis is ∼6°–10° lower than that back-analyzed for the stability of this event. Thus, a mechanism to reduce strength along the rupture surface immediately following the initial instability was required for catastrophic failure of this event. This mechanism is poorly understood at present, but is likely crucial for understanding the transition from an initially stable slope to a catastrophic, long runout rock avalanche., Frontiers in Earth Science, 8, ISSN:2296-6463

Mapping landslides using automated methods is a challenging task, which is still largely done using human efforts. Today, the availability of high-resolution EO data products is increasing exponentially, and one of the targets is to exploit this data source for the rapid generation of landslide inventory. Conventional methods like pixel-based and object-based machine learning strategies have been studied extensively in the last decade. In addition, recent advances in CNN (convolutional neural network), a type of deep-learning method, has been widely successful in extracting information from images and have outperformed other conventional learning methods. In the last few years, there have been only a few attempts to adapt CNN for landslide mapping. In this study, we introduce a modified U-Net model for semantic segmentation of landslides at a regional scale from EO data using ResNet34 blocks for feature extraction. We also compare this with conventional pixel-based and object-based methods. The experiment was done in Douglas County, a study area selected in the south of Portland in Oregon, USA, and landslide inventory extracted from SLIDO (Statewide Landslide Information Database of Oregon) was considered as the ground truth. Landslide mapping is an imbalanced learning problem with very limited availability of training data. Our network was trained on a combination of focal Tversky loss and cross-entropy loss functions using augmented image tiles sampled from a selected training area. The deep-learning method was observed to have a better performance than the conventional methods with an MCC (Matthews correlation coefficient) score of 0.495 and a POD (probability of detection) rate of 0.72 ., Remote Sensing, 12 (3), ISSN:2072-4292

DECOVALEX 2019 Symposium Abstracts