Role of measured and simulated water content patterns for landslide early warning systems

Landslide early warning systems based on rainfall intensity and duration thresholds neglect the role of antecedent rainfall events on the hydration state that defines the disposition of a steep slope to fail in forthcoming rainfall events. Water content, water potential and mechanical strength of the soil largely depend on the antecedent rainfall signature and the soil hydraulic properties. To investigate how soil moisture information can be used for LEWS, six soil moisture measuring stations have been installed in the Napf-Emmental region as part of an ongoing pilot study to develop a territorial LEWS in Switzerland. In order to estimate the spatial distribution of the initial water content and its effect on landslide frequency and magnitude, we combine water content patterns from these stations, topographic disposition and regional rainfall data. The calculated soil water content patterns are used as input for landslide triggering simulations using the hydromechanical model framework STEP-TRAMM. STEP-TRAMM calculates the load distribution between mechanically interacting soil columns that may result in progressive failure culminating in hazardous landslides. Using landslide inventory data for the pilot region, we calibrate and validate the landslide model and evaluate the role of uncertainty in initial water content pattern on landslide characteristics and rainfall thresholds. We found high correlations between the measured and simulated water content based on rainfall characteristics and topographic disposition (R² = 0.94), allowing a reasonable estimate of the spatial distribution of the initial water content which underlines the outcome of further landslide triggering simulations.