Andrea Pedrazzini

Machine Learning Feature Selection Methods for Landslide Susceptibility Mapping

This paper explores the use of adaptive support vector machines, random forests and AdaBoost for landslide susceptibility mapping in three separated regions of Canton Vaud, Switzerland, based on a set of geological, hydrological and morphological features. The feature selection properties of the three algorithms are studied to analyze the relevance of features in controlling the spatial distribution of landslides. The elimination of irrelevant features gives simpler, lower dimensional models while keeping the classification performance high. An object-based sampling procedure is considered to reduce the spatial autocorrelation of data and to estimate more reliably generalization skills when applying the model to predict the occurrence of new unknown landslides. The accuracy of the models, the relevance of features and the quality of landslide susceptibility maps were found to be high in the regions characterized by shallow landslides and low in the ones with deep-seated landslides. Despite providing similar skill, random forests and AdaBoost were found to be more efficient in performing feature selection than adaptive support vector machines. The results of this study reveal the strengths of the classification algorithms, but evidence: (1) the need for relying on more than one method for the identification of relevant variables; (2) the weakness of the adaptive scaling algorithm when used with landslide data; and (3) the lack of additional features which characterize the spatial distribution of deep-seated landslides.

Preliminary Slope Mass Movement Susceptibility Mapping Using DEM and LiDAR DEM

Hazard mapping in mountainous areas at the regional scale has greatly changed since the 1990s thanks to improved digital elevation models (DEM). It is now possible to model slope mass movement and floods with a high level of detail in order to improve geomorphologic mapping. We present examples of regional multi-hazard susceptibility mapping through two Swiss case studies, including landslides, rockfall, debris flows, snow avalanches and floods, in addition to several original methods and software tools. The aim of these recent developments is to take advantage of the availability of high resolution DEM (HRDEM) for better mass movement modeling. Our results indicate a good correspondence between inventories of hazardous zones based on historical events and model predictions. This paper demonstrates that by adapting tools and methods issued from modern technologies, it is possible to obtain reliable documents for land planning purposes over large areas.

Structural analysis of Turtle Mountain: origin and influence of fractures in the development of rock slope failures

Abstract Large slope failures in fractured rocks are often controlled by the combination of pre-existing tectonic fracturing and brittle failure propagation in the intact rock mass during the pre-failure phase. This study focuses on the influence of fold-related fractures and of post-folding fractures on slope instabilities with emphasis on Turtle Mountain, located in SW Alberta (Canada). The structural features of Turtle Mountain, especially to the south of the 1903 Frank Slide, were investigated using a high-resolution digital elevation model combined with a detailed field survey. These investigations allowed the identification of six main discontinuity sets influencing the slope instability and surface morphology. According to the different deformation phases affecting the area, the potential origin of the detected fractures was assessed. Three discontinuity sets are correlated with the folding phase and the others with post-folding movements. In order to characterize the rock mass quality in the different portions of the Turtle Mountain anticline, the geological strength index (GSI) has been estimated. The GSI results show a decrease in rock mass quality approaching the fold hinge area due to higher fracture persistence and higher weathering. These observations allow us to propose a model for the potential failure mechanisms related to fold structures.

From the source area to the deposit: Collapse, fragmentation, and propagation of the Frank Slide

The combination of structural data from the source area and descriptive data from the deposit’s carapace, as well as remote sensing and statistical analysis, allows a better understanding of the collapse, fragmentation, and propagation processes of the Frank Slide rock avalanche. The in situ observed conservation of the stratigraphic sequence of the Turtle Mountain anticline’s normal limb in the deposit is interpreted as the consequence of the collapse mode, involving simple shearing of the mass accompanied by a rotational movement, i.e., a “simple shear” parallel to the topography, and a breakage of the hinge, followed by overthrusting of the normal limb on the inverse limb. Coherence between the block size distributions of the source area and the deposit carapace is interpreted as a demonstration of the primary control of preexisting fracturing on the fragmentation processes. Remote-sensing indexes allow us to uncover a priori hidden morphological features preserved on the surface of the deposit, i.e., longitudinal and compressional features, as well as lobes. Their location on the carapace provides evidence of lateral heterogeneity in the propagation, highlighted by three different types of propagation behaviors. This comprehensive study not only provides elements that contribute to an understanding of the Frank Slide rock avalanche, but it also provides insight into essential parameters to take into account in further modeling of these types of phenomena.

Inventory of Rock Slope Deformations Affecting Folded Sedimentary Layers in Moderate Relief Context: The Case of the Livingstone Range Anticlinorium, AB, Canada

This contribution investigates the distribution of rock slope deformations (RSD) and their relationships with tectonic structures, lithological and topographic contexts. This is performed in a moderate relief context (1,100–2,500 masl.) affected by thin-skin tectonic. The proposed inventory, along the Livingstone range anticlinorium (LRA), contains 160 gravitational events that are spatially clustered in four areas. The thrust-folds properties (density, geometry and lithology) and the local relief appear to primary influence the development of RSD in this context.