Rainfall Induced Slope Reliability Analysis Using Radial Degraded Random Fields

A radial degradation model is originated and integrated with the conventional stationary Gaussian random field to model the spatial variability of soil properties under rainfall conditions. Random forest surrogate model is combined with Monte Carlo Simulation to conduct rainfall induced transient slope reliability analysis. The proposed methodology is illustrated through a reservoir slope. The effects of rainfall conditions, degradation coefficient (DB), degradation model, and the vertical scale of fluctuation on the slope transient reliability are investigated in detail. Simulation results show that as DB decreases from 1.0 to 0.75, the influence of rainfall intensity on the deterioration of slope stability level turns to be weak, and the most significant day tends to move forwards. The advanced rainfall pattern will significantly deteriorate the slope stability level at the first and second day as compared with uniform rainfall pattern. The most significant effect of DB on the deterioration of slope stability level is found at different rainfall conditions as the rainfall evolves. The choice of rainfall degradation model significantly affects the prediction of slope failure probability. The slope failure probability exhibits an increase trend as the vertical scale of fluctuation increases (i.e., the spatial variability grows insignificant).