Effects of soil spatial variability at the hillslope and catchment scales on characteristics of rainfall‐induced landslides

Abstract

Spatial variations in soil properties affect key hydrological processes, yet their role in soil mechanical response to hydro-mechanical loading is rarely considered. This study aims to fill this gap by systematically quantifying effects of spatial variations in soil type and initial water content on rapid rainfallinduced shallow landslide predictions at the hillslope-and catchment-scales. We employed a physicallybased landslide triggering model that considers mechanical interactions among soil columns governed by strength thresholds. At the hillslope scale, we found that the emergence of weak regions induced by spatial variations of soil type and initial water content resulted in early triggering of landslides with smaller volumes of released mass relative to a homogeneous slope. At the catchment scale, initial water content was linked to a topographic wetness index, whereas soil type varied deterministically with soil depth considering spatially correlated stochastic components. Results indicate that a strong spatial organization of initial water content delays landslide triggering, whereas spatially linked soil type with soil depth promoted landslide initiation. Increasing the standard deviation and correlation length of the stochastic component of soil type increases landslide volume and hastens onset of landslides. The study illustrates that for similar external boundary conditions and mean soil properties, landslide characteristics vary significantly with soil variability, hence it must be considered for improved landslide model predictions.

The role of uncertainty in bedrock depth and hydraulic properties on the stability of a variably-saturated slope

“…For instance, some authors have pointed out that landslides are controlled by critical rainfall intensity and duration [7], anisotropic hydraulic conductivity [7,58], slope geometry [2][3][4][5][6], initial saturation [3,4,7], recharge from bedrock [3,58], and strength parameters [11] on landslides. These studies, however, have been restricted to simplified geometries.…”

“…These two constitutive relationships generally exhibit considerable spatial variability, and are very time consuming, labor intensive, and costly to measure at the scale of a hillslope [22,23]. Whereas several authors have investigated the effect of the saturated hydraulic conductivity on slope (in) stability [12,[24][25][26], relatively few contributions in the geotechnical literature have explored properly the coordinated impact of SWCC and HCF uncertainty on the SF values derived from slope stability studies [4,6,18,27]. We agree wholeheartedly with Liang and Uchida [23] that a detailed characterization of the temporal and spatial variability of the moisture content of the soil mantle is warranted in slope stability studies.…”

“…As hillslope interiors are costly, labor-intensive, and difficult to access, most slope stability studies use a relatively simple description of the bedrock surface [3,6,7,11]. Resulting topographic maps often do not do justice to complex field measurements of the bedrock depth, which often demonstrate significant spatial variability [3,33] with a geometry that is difficult to characterize adequately with some closed-form mathematical expression, while hydraulic and strength parameters can vary abruptly at the soil-bedrock interface.…”

“…Numerous contributions to the geotechnical literature have investigated the triggering mechanisms and probability of occurrence of rainfall-induced landslides. This body of work has shown that the stability of hillslopes is controlled by many different factors including surface topography [2], depth to unweathered bedrock [3][4][5][6], the hydraulic properties [7,8] and shear strength [9][10][11] including their spatial variability [12,13].…”

Probabilistic landslide ensemble prediction systems: lessons to be learned from hydrology

“…Probabilistic approaches allow for a more thorough consideration of uncertainties and inherent variability in model-specific parameters. Spatially varying parameters (both geotechnical and hydraulic) are usually represented as univariate distributions of random variables based on an underlying pdf and statistical characteristics (Fan et al, 2016). Friction angle and cohesion are commonly considered to be such varying variables that are treated in a probabilistic way for model parameterization (e.g., Park et al, 2013;Chen and Zhang, 2014;Raia et al, 2014;Salciarini et al, 2017).…”