Quantifying the three-dimensional effects of anisotropic soil horizons on hillslope hydrology and stability

Rainfall-induced shallow landslides cause significant damage involving loss of life and property. Many hydrological processes such as rainfall infiltration, soil water dynamics, and slope stability are controlled by unsaturated soil properties, such as unsaturated hydraulic conductivity. Natural soils often exhibit a certain degree of anisotropy in hydraulic conductivity due to stratification and compaction mechanisms in soil formation processes. In this paper we investigate the effect of soil hydraulic conductivity anisotropy (SHCA) on hillslope hydrology and stability using a three-dimensional hydrological model coupled with a probabilistic infinite slope stability model. The model is applied in two independent case studies. The first aims to quantify the combined effect of different anisotropy ratios (lateral/normal saturated hydraulic conductivity) and hillslope morphologies (convex, concave, and planar) on slope stability. Anisotropy ratios are assumed in this case higher than one (1, 2, 10). Results show that increasing the anisotropy ratio (from 1 to 10) anticipates the failure time (from 12 to 9 h after the start of rainfall) and that in concave morphologies the unstable area tends to be wider than planar and convex. The second application aims to simulate the soil moisture dynamic and the probability of failure at different depths (100, 500, and 900 mm) of a stratified volcanic soil, making leverage on the model flexibility to accommodate SHCA. No assumptions are made on the anisotropy ratio in this case study. Our results, based on model parameter calibration and verification against in-situ soil moisture measurements during the year 2009, showed good model performance in simulating the soil moisture dynamic (Kling-Gupta Efficiency higher than 0.78) and confirmed no failure for the simulated year. The promising results support the aspiration that the physically based hydrological model can complement and improve the current predictions of landslide early warning systems.