Spatially distributed modeling of rainfall-induced landslides in shallow layered slopes

The zoning of landslide susceptibility on layered landscapes is a key challenge for regional hazard analyses. From a modeling standpoint, the combination of transient infiltration and vertical heterogeneity can lead to hydro-mechanical processes that are difficult to incorporate in spatially distributed frameworks. In this work, a physically based model for the efficient generation of regional landslide susceptibility maps in layered landscapes is presented. The formulation involves the discretization of a digital terrain into slope units, thus enabling the incorporation of georeferenced datasets to define the input variables. Model computations rely on a vectorized finite element (FE) solver that performs simulations of vertical unsaturated flow and slope stability analyses. The framework allows the use of different meshes across the region to efficiently allocate the computational cost associated with layers of variable thickness and/or complex stratigraphy. The model is used to analyze a series of documented shallow landslides that occurred in a region covered by stratified volcanic deposits. In addition to the simulation of layered profiles constrained by field and laboratory data, two simplified scenarios are considered in which homogeneous slopes with different values of hydraulic conductivity K _s are used. It is shown that, while the homogeneous models may have an acceptable spatial performance in some sectors of the landscape, the use of homogenized values of K _s leads to inconsistent temporal sequences of landslide triggering, as well as to failure depths always located at the base of the slope. By contrast, the use of stratified profiles leads to an improved spatiotemporal performance over the whole region, as well as to computed failure depths that are consistent with landslide inventories. The proposed methodology provides a useful tool for landslide hazard studies in that it not only addresses the computational challenges associated with multiple slope stability analyses, but it also enables the incorporation of system properties that are often neglected in spatially distributed modeling frameworks.