Probabilistic hazard assessment of landslide-induced river damming

Abstract

Landslide-induced river damming poses a considerable threat to the safety of humans and infrastructure. Prediction of landslide-induced river damming is of great significance for quantitative risk assessment and emergency response planning. However, owing to the large uncertainties embedded in the input parameters and dynamic numerical models, a reliable physically based prediction of landslide dam formation is still challenging. In this study, we proposed a probabilistic framework to predict rockslide-induced river damming and its associated barrier lake. Both parameter and model uncertainties were considered to reproduce the run-out process and the deposition behavior of the Baige landslide based on dynamic numerical simulation, while calibrating the input parameters through a sequential Bayesian back analysis. The first slide of the Baige landslide was considered to calibrate the input parameters with observations at several deposition depths. The proposed method was validated by predicting the second slide-induced river damming using the calibrated results from the first slide. Furthermore, the second slide with deposition depth observations was employed to update the input parameters again. Through the sequential Bayesian back analysis, the probability of the river damming induced by the potentially unstable rock masses was predicted, which yielded a hazard zonation map of the barrier lake exceeding various water levels. This hazard zonation map may be employed to guide quantitative risk assessment and corresponding emergency response plans for future landslide-induced upstream backwater-inundation.