PROBABILISTIC PREDICTION OF SLOPE FAILURE TIME BASED ON DISPLACEMENT

Displacement is a significant apparent feature of the slope stability. The Log-fit method is a common method to predict the time of the slope failure based on the observed displacement data. In practice, the influence of measurement error, environmental noise and model assumptions may make it difficult to predict the slope failure time. In this paper, the uncertainties in slope failure time prediction were divided into observational uncertainty and model uncertainty. We suggested a method to calibrate the observational uncertainty based on the residual Bootstrap method. We collected the cases of rainfall induced landslides, and then calibrated the model uncertainty of the Log-fit method when adopted to predict the failure time of rainfall induced landslides. On this basis, we proposed a probabilistic method to predict the slope failure time considering both uncertainties. The suggested method can provide both the most likely failure time of the slope and the failure probability associated with different predicted failure time. It is demonstrated through the illustrative examples that the result from the probabilistic method accords better with the observed failure time of the slope, which can provide useful information for the early warning of landslide disasters.