An Efficient Method for Calculating Annual Probability of Rainfall-Induced Slope Failure

Abstract:

Calculating the annual probability of slope failure is crucial for landslide risk assessment. In this paper， an efficient method for calculating the annual probability of slope failure under rainfall conditions is proposed considering both uncertainties of rainfall and soil parameters. First， according to the theory of seepage and strength of unsaturated soil， the limit equilibrium method is used to obtain safety factors of the slope under a given rainfall event. Then， the slope stability analysis is regarded as a binary classification problem， and the support vector machine is used to build a surrogate model to predict the slope failure. With the surrogate model， the failure probabilities of the slope under different rainfall conditions are analyzed to construct the fragility surface. The annual most dangerous rainfall is identified based on the maximum slope failure probability. Finally， according to the total probability theorem， the Monte Carlo simulation is used to estimate the average failure probability of a slope under the annual most dangerous rainfall， i.e.， the annual probability of rainfall-induced slope failure. The proposed method is applied to study the annual probability of rainfall-induced slope failure in a certain region. The results show that the probability of slope failure for a given rainfall intensity increases nonlinearly with rainfall duration. As the rainfall intensity increases， the probability of slope failure increases. The annual most dangerous rainfall is more intense and longer in duration than normal rainfall but has a lower likelihood of occurrence. The annual probability of slope failure is proportional to the average number of rainfall events per year and is more sensitive to effective cohesion and friction angle. The method proposed in this paper can provide an efficient tool for rainfall-induced landslide risk assessment.